- Email: [email protected]
Reproducibility of characteristic marks on fired cartridge cases from five Chinese Norinco QSZ-92 9 × 19 mm pistols
Forensic Science International 278 (2017) 78–86
Contents lists available at ScienceDirect
Forensic Science International journal homepage: www.elsevier.com/locate/forsciint
Reproducibility of characteristic marks on fired cartridge cases from five Chinese Norinco QSZ-92 9 19 mm pistols Kaifeng Zhanga , Yaping Luob,* , Peng Zhouc a b c
School of Forensic Science, People’s Public Security University of China, No.1, Muxidi South Street, Xicheng District, Beijing 100038, China Graduate School, People’s Public Security University of China, No.1, Muxidi South Street, Xicheng District, Beijing 100038, China School of Combat Skills and Police Tactics, People’s Public Security University of China, No.1, Muxidi South Street, Xicheng District, Beijing 100038, China
A R T I C L E I N F O
Article history: Received 28 December 2016 Received in revised form 3 May 2017 Accepted 27 June 2017 Available online 4 July 2017 Keywords: Firearms identification Evofinder Ballistic database Automatic correlation
A B S T R A C T
Reproducibility of characteristic marks on fired cartridge cases from five Chinese Norinco QSZ-92 9 19 mm pistols over the course of 3070 shots is addressed using an Evofinder1 system. The first 20 cartridges are all studied, while one random sample out of every ten consecutively fired cartridge cases is studied for the rest 3050 rounds. As such, a total of 325 cartridges are entered into the system for each pistol and a database consisting of 1625 cartridges from five pistols is established in this paper. Both onscreen examination and automatic correlation are used to examine the reproducibility. In the onscreen examination process, it is possible for firearm examiners to positively identify the first cartridge to the 500th, 1000th, 1500th, 2000th, 2500th, 3000th, 3070th cartridge. In the automatic correlation process, the first cartridge of each firearm is separately performed correlation against the database. It is found that the similarity score for match objects changes slightly as the shot number increases, indicating slight variations of marks. However, these variations have not prohibited the Evofinder1 system to make correct correlation and rank 100% of the known match objects on the top of the correlation list by either firing pin impression or breech face mark. Thus, the reproducibility is proved statistically and objectively in this paper. © 2017 Elsevier B.V. All rights reserved.
1. Introduction Fired cartridge cases are used to identify the specific firearm that fired them, based on firing pin impressions, breech face marks, ejector marks and many other characteristic marks left by the firearm during firing process. Firing pin impressions are indentations created when the firing pin of a firearm strikes the primer of cartridge case. If the nose of the firing pin has manufacturing imperfections or damages, these potentially unique characteristics can be impressed into the metal of the primer by the strike. The breech face leaves its characteristic marks when the head of a cartridge case slams against the breech face due to the pressure developed by the combustion gases within the cartridge. Breech marks can be found as parallel breech marks, circular breech marks or no obvious pattern, just as a stippled or granular appearance. Ejector marks are generated when cartridge cases are ejected from the action of a firearm. All of these above marks can be used for firearms identification.
* Corresponding author. Tel.: +86 10 83903102; fax: +86 10 83903102. E-mail address: [email protected] (Y. Luo). http://dx.doi.org/10.1016/j.forsciint.2017.06.032 0379-0738/© 2017 Elsevier B.V. All rights reserved.
However, these marks may be affected by the change of the characteristics on the firing pin, breech face and ejector owing to the use of the firearm. The influence of the use of firearm on the reproducibility of characteristic marks on cartridge cases has been researched in numerous articles [1–10]. In general, no significant changes have been observed and it is possible for firearm examiners to identify the first and last fired cartridges as to be fired in the same firearm [5]. In previous researches, it is mainly by microscopic comparison that firearm examiners examine cartridges fired in the same firearm, which is an extremely time-consuming process. Therefore, only a few number of samples of cartridges of different shot numbers could be compared. Besides, it strongly depends on firearm examiners’ subjective interpretation whether sufficient agreement exists between the first and later fired cartridges or not. In a paper by Grom and Demuth, an IBIS BrassTRAX-3DTM System was employed to examine the reproducibility [10]. A GlockTM pistol was shot for 500 firings and cartridge numbered 1, 10, 25, 50, 100, 150, 200, 250, 300, 350, 400, 450, 500 were recovered and introduced into the system. The first cartridge was retained in a database which contain other 908 cartridges that bear the similar class characteristics as a known test, and then the later
K. Zhang et al. / Forensic Science International 278 (2017) 78–86
fired cartridges were correlated against the database. Consequently, the first cartridge was successfully found as a “hit” for all twelve cartridge samples, demonstrating that the individual characteristics did not change significantly throughout 500 firings, which made the IBIS1 system could correctly associated them. In addition, a trend of minor drop of firing pin impression rankings was observed while breech face rankings remained consistently near the top as the shot number grew. To survey the reproducibility of characteristic marks on fired cartridge cases from five Chinese Norinco QSZ-92 9 19 mm pistols, a method similar to Grom’s is used in this study. The reproducibility is examined using an Evofinder1 system, which is developed by ScannBI Technology. Both onscreen examination and automatic correlation examination are applied, especially the latter. In the correlation process, a large number of cartridges fired from the same firearm are automatically compared by the Evofinder1 system, where the reproducibility and variation of marks are presented by a similarity score objectively. 2. Materials and methods Firearms used to perform the test are five QSZ-92 9 19 mm pistols designed by Norinco. As shown in Table 1, these firearms will be called A, B, C, D and E, respectively, throughout the paper. The ammunition is DAP92-9 ammunition which is also manufactured by Norinco. A total of 3070 rounds were fired for each firearm and the firearms had not been cleaned during the test. Firearms were fired in a cleared area and the cartridge cases were recovered in the following way. The first 20 cartridge cases fired by the firearms were collected and saved in individual envelopes. From the 21st to the 3070th round, every ten consecutively fired cartridge cases were recovered and placed into one envelope together, thus fire sequence was not differentiated any more among these ten cartridges. Firing date, firearm name and number of shots were recorded on the envelope. To facilitate the project, 3070 cartridge cases fired from each firearm were not all used for further study. The first 20 cartridge cases were all studied. For the rest 3050 cartridge cases, it was chosen to select one random sample out of ten consecutive cartridge cases in the very envelope as mentioned. The selected cartridge cases were studied as the 30th, 40th, . . . , 3070th round. Therefore, 325 cartridge cases were selected for each firearm, and 1625 cartridge cases were retained for study in total. The Evofinder1 system applied in this paper is composed of a data acquisition station (DAS), a specimen analysis system (SAS) with the software version 6.3 and an expert working station (EWS). DAS is aimed at digitally recording objects surfaces and saving them into the database. SAS is aimed at performing an automatic identification of the object under examination against objects in the database and storing high-quality digital images with corresponding information. EWS is aimed at onscreen identification as well as sending a request to SAS for automatic correlation procedure start. EWS provides with hit-lists of objects previously saved in the database and similar to the object under examination [11]. Objects in the database are ranked according to a similarity score ranging from 0 to 1, which is automatically calculated on the Table 1 Five firearms used to perform the test firing. Firearm
Serial number
A B C D E
012160 007252 012375 012624 010457
79
basis of system’s algorithm. Higher the similarity score is, the object in the database looks more similar to the object under examination. The selected 1625 cartridge cases were entered into the Evofinder1 system under its manual protocol. We cleaned, oriented and scanned all of the cartridges in the same way. It was decided to orient the ejector mark at 9 o’clock position. Cartridges were named by a combination of pistol name and shot number in the database. For example, A0001 represents the first cartridge from pistol A, B0500 means one of the 491th–500th cartridge cases from pistol B and so forth. As such, a database of 1625 cartridge cases which could be equally divided into five groups according to the source was established. Afterwards, onscreen examination of cartridge cases was conducted. Images of three-dimensional information of the cartridge markings were compared side by side on the screen. For each firearm, firstly the beginning 20 cartridges were intercompared to build the elements of identity and secondly the 500th, 1000th, 1500th, 2000th, 2500th, 3000th and 3070th cartridges were separately compared to the first cartridge. In the procedure of automatic correlation, the first cartridge of each firearm (A0001, B0001, C0001, D0001 and E0001) was run correlation against the database separately. It should be noted that when one of the above cartridge cases is used for search, the cartridge itself will not be correlated while other 1624 cartridges are all correlated. On starting a correlation, we chose to search by firing pin impression and breech face mark and abandoned ejector mark because of its poor quality. The optional function of “Breech nonoriented”, which is usually used in situation where cartridge is scanned without orientation, was not selected in this study. 3. Results and discussions 3.1. Onscreen examination 3.1.1. Preliminary examination There exist several marks on the bottom of fired cartridge case, including class characteristics and individual characteristics. The class mark displaying an interesting horseshoe shape at the head of the cartridge is impressed by the breech bolt. It is called “horseshoe mark” by Chinese firearm examiners. As to individual mark, the firing pin impression shows ordinary circle shape while the flow back mark surrounding it looks extraordinarily. The breech face mark is found as normal arc and line. The general appearance of a fired cartridge case is illustrated in Fig. 1. 3.1.2. Identity elements By inter-comparing images of the first 20 cartridges, the identity elements of marks for each firearm was established. It is found that five pistols produce similar class characteristics as well as dissimilar individual characteristics. Although the class characteristics are extremely alike, minute differences can still be observed. For example, the horseshoe mark, which is a typical class characteristic, appears almost the same on cartridges from different guns, yet the curvature of the horseshoe shape differs. As to the individual characteristics, either firing pin impression or breech face mark can be used to identify these cartridges to one of the five firearms separately. For firing pin impression, not only the tiny marks lying at the bottom of the aperture but the flow back marks are different as well. Breech face mark also shows distinctive for each firearm. 3.1.3. Reproducibility Comparing the first cartridge to the 500th, 1000th, 1500th, 2000th, 2500th, 3000th, 3070th cartridge, characteristics on the first cartridge can also be observed on the later fired cartridge
80
K. Zhang et al. / Forensic Science International 278 (2017) 78–86
Fig. 1. Image on the left indicates the “horseshoe mark” by a yellow dotted line and the flow back mark by a blue solid line. The same image without annotations is on the right in comparison.
samples, which means characteristics are reproducing. As the shot number increases, slight variation of the characteristics is noted for both firing pin impression and breech face mark. Images in Fig. 2 indicate firing pin impression and breech face mark of the 1st, 1000th, 2000th and 3000th cartridge of each firearm. 3.2. Automatic correlation
study. For breech face mark, the variation of the similarity score conveys interesting information. There may be some special characteristics on breech faces of the five new firearms. These characteristics present irregularly from round to round and disappear completely after tens of firings. We guess that these special characteristics may be caused by some substance existing in breech face of new weapons, thus they are not the real characteristics of the firearm. To check this hypothesis, cartridges of other shot number can be used to perform a correlation in future studies.
When the first cartridge of each firearm is used for search in the database, cartridges with the same source and other four sources are associated equally. To some extent, cartridges fired in other four firearms act as background interference. It should be specified that the horseshoe mark and the flow back mark mentioned in Section 3.1 do not take part in the automatic correlation in the light of system’s algorithm.The correlation results are given in two separate lists: a list of firing pin impression (FPI) and a list of breech face mark (BF).
3.2.2. Distribution of similarity score In this section, the similarity score of both the known match objects and the known non match objects in each correlation list is examined. Using the Minitab1 17 software, the density distribution histograms are charted and the descriptive statistics are calculated.
3.2.1. Similarity score as a function of shot number In theory, if a cartridge is associated with many cartridges fired by the same firearm, the similarity score should be the same, in the case of an ideal reproducibility and the most effective automated ballistic identification system. However, there does not exist an ideal reproducibility, so it is reasonable to get different similarity scores. In this paper, we are interested in verifying if and how the similarity score varies as the shot number increases for each firearm. With the help of filter function, it is easy to exclusively show similarity scores of the known match objects. The similarity score as a function of shot number is examined for firing pin impression and breech face mark separately, as shown in Figs. 3 and 4. In Fig. 3, we can notice a common sign that as the shot number increases, more dots appear at lower units. It demonstrates that the similarity score decreases as the shot number increases. On the contrary, the similarity score for breech face mark is of interest at the very beginning of the shot number, as illustrated in Fig. 4. It is found that the similarity score ranges so rapidly that the dots arrayed almost in a straight line along the y-axis. Towards right along the x-axis, the similarity score keeps in a relatively low level and no regular trend is observed again. The change of the similarity score reveals that firearm signatures evolve over the course of 3070 shots. This parallels the results of onscreen examination. With firing repeatedly, the firing pin of the firearm wear and then the change of firing pin impression on fired cartridges results. Supposing more cartridges are fired in these five guns, a more obvious diminution of the similarity score may be observed, which are not undertaken in this
3.2.2.1. Firing pin impression. Histograms for firing pin impression correlation results are illustrated in Fig. 5. We can learn three interesting things from them. Firstly, the bars representing match objects are built in higher similarity score area while non match objects in lower similarity score area. As can be seen in five histograms, the distance between match and non match objects is notable. This result shows that Evofinder1 has correctly calculated with higher similarity score for known match objects than known non match objects, which indicates a good correlation performance of Evofinder1. Secondly, the distributions of the similarity score for the non match objects are extremely narrow. In contrast, the distribution of the match objects is wide. To explain this situation, we have specially queried personals of Evofinder1 company. In principle, when two cartridges bear more similar areas, the similarity score given by the Evofinder1 system will be higher. Imagine an ideal firearm which exactly reproduces the same imprint of its individual characteristics on every round, and thus any two cartridges fired in this firearm will be associated with similarity score close to 1. Likewise, imagine ideal firearms which own absolutely different individual marks, and then any two cartridges fired in different firearms will be associated with similarity score close to 0. In such perfect scenario, distribution of matches will present high mean value near 1 with low deviation and distribution of non matches will present low mean value near 0 with low deviation too. However, real firearms are not so perfect that neither the reproducibility nor the individuality is absolute, because of which both match distribution and non match distribution are wider.
K. Zhang et al. / Forensic Science International 278 (2017) 78–86
81
Fig. 2. Images are grouped as the following rule: the x-axis represents the shot number and the y-axis represents the letter of weapon.
For match objects, reproduced imprints are sometimes located at the same parts of cartridge as a result of relatively high similarity score (for example 0.8), and sometimes in different parts of cartridge with small intersection area as a result of relatively low
similarity score (for example 0.4 or 0.5). As such, a deviation of the similarity score of match objects can be imagined. For non match objects, though the similarity score will not be equal too, nevertheless, they are usually very low (typically lower than
82
K. Zhang et al. / Forensic Science International 278 (2017) 78–86
Fig. 3. Graphs representing the similarity score as a function of the shot number. The similarity score in A0001 FPI, B0001 FPI, C0001 FPI, D0001 FPI and E0001 FPI stands for the known match objects in the FPI correlation results of A0001, B0001, C0001, D0001, E0001 correspondingly.
0.15), resulting in a smaller deviation in comparison to match objects. Thus, it is logical that the distribution of matches is wide while the distribution of non matches is narrower in this paper. Reproducibility and individuality are correctly presented in this correlation result, which is another indicator of the good correlation performance of Evofinder1. In real cases, a narrow distribution of non match objects is needed to improve the probability of match object to be found. The narrower the distribution of non match objects, the probability is less that some insolent non match object will override the similarity score of match objects and kick out the match objects from the top of the ranking list. Thirdly, in all five histograms, the distribution of match objects shows an interesting tail towards lower similarity scores, which indicate that the data is skew to the left. Furthermore, the width of the match data for each firearm (full width at half maximum or standard deviation) appears approximately equal except for pistol B.
In Table 2, the descriptive statistics: skewness, standard deviation and mean for each firearm in each histogram in Fig. 5 are summarized. These data are meaningful, especially the numbers in bold type which stand for the values of match objects. It is found that the absolute value of numbers in bold is maximal in both the row and the column it locates for each statistics. This result is in consistent with the information in histograms. In addition, the negative value of skewness of for match objects has responded to the “tails” towards left in histograms, disclosing the decline trend of the similarity score, wherein the evolution of firing pin impression during extensive shots is further proved. Five pistols have approximately equal standard deviation values and mean values (numbers in bold) except for pistol B, which gives the lowest value of both statistics. The difference in mean value may be explained by the fact that similarity score is determined by the quality of mark. For instance, it is clear that E0001 bears better quality of firing pin impression than B0001 from Fig. 2, and hence the mean value for pistol E is higher than that for pistol B.
K. Zhang et al. / Forensic Science International 278 (2017) 78–86
83
Fig. 4. Graphs representing the similarity score as a function of the shot number. The similarity score in A0001 BF, B0001 BF, C0001 BF, D0001 BF and E0001 BF stands for the known match objects in the BF correlation results of A0001, B0001, C0001, D0001, E0001 correspondingly.
Furthermore, it seems that the lowest mean value has led to the lowest standard deviation for pistol B, which discloses the correlation between mean and standard deviation to some extent. However, more evidence is required to prove it. 3.2.2.2. Breech face mark. The same histograms are made for breech face mark, as shown in Fig. 6. There is a tiny overlap between the bars of match objects and non match objects in all five histograms. Still, the distribution of non match objects is very narrow as opposed to match objects with a wide distribution. Five firearms show about equal width for the distribution of match objects except for pistol C, the data of which looks a bit narrow. Also, the values of skewness, standard deviation and mean of every firearm in each histogram are summarized, as shown in Table 3. The behavior that the value representative of match objects (numbers in bold) is maximal takes place in three statistics apart
from skewness, the value of which has not showed any regularity. Again, for the values of standard deviation and mean (numbers in bold), five firearms show about equal numbers. It is pistol C that gives the lowest value of both statistics this time. The relatively less outstanding characteristics on breech face mark of pistol C, as shown in Fig. 2, can be used to explain the lowest mean value. What is interesting is the assumption that the lowest mean leads to the lowest standard deviation works here. Nevertheless, we can not draw a certain conclusion and more data are needed to check the correlation between mean and standard deviation. 3.2.2.3. Comparison of descriptive statistics for FPI and BF. To compare the statistics value of match objects for two types of marks: firing pin impression and breech face mark, the numbers in bold in Tables 2 and 3 are grouped in Table 4. In the column of skewness, it is found that the values are negative for firing pin impression as opposed to positive for breech
84
K. Zhang et al. / Forensic Science International 278 (2017) 78–86
Fig. 5. Histograms in A0001 FPI, B0001 FPI, C0001 FPI, D0001 FPI and E0001 FPI representing the density distribution of the similarity score under examination of A0001, B0001, C0001, D0001 and E0001 by firing pin impression correspondingly.
face mark. The variation of breech face mark has not been indicated by skewness as clear as that of firing pin impression. For each firearm, the standard deviation value of two types of marks is approximately equal, which indicates that the extent of the reproducibility of firing pin impression and breech face mark on
the first cartridge is at the same level. With regard to the mean value, firing pin impression presents higher value than breech face mark. It may be caused by the different quality of two types of marks on the first cartridge, as well as the difference in respective algorithms. According to the personals of Evofinder1 company, the
Table 2 Summary of descriptive statistics for each firearm in each histogram of firing pin impression. Specimen under examination
Descriptive statistics Skewness
A0001 B0001 C0001 D0001 E0001
Standard deviation
Mean
A
B
C
D
E
A
B
C
D
E
A
B
C
D
E
0.69 0.06 0.36 0.10 0.11
0.27 0.84 0.51 0.03 0.07
0.10 0.60 1.10 0.09 0.06
0.10 0.37 0.21 1.31 0.09
0.08 0.42 0.16 0.11 1.52
0.06 0.01 0.01 0.01 0.01
0.01 0.02 0.01 0.01 0.01
0.01 0.01 0.05 0.01 0.02
0.01 0.01 0.01 0.05 0.02
0.01 0.01 0.01 0.02 0.05
0.47 0.07 0.09 0.08 0.09
0.08 0.19 0.06 0.08 0.09
0.08 0.07 0.36 0.09 0.09
0.09 0.07 0.09 0.37 0.10
0.09 0.06 0.09 0.10 0.53
K. Zhang et al. / Forensic Science International 278 (2017) 78–86
85
Fig. 6. Histograms in A0001 BF, B0001 BF, C0001 BF, D0001 BF and E0001 BF representing the density distribution of the similarity score under examination of A0001, B0001, C0001, D0001 and E0001 by breech face mark correspondingly.
algorithms for breech face and firing pin impression are different, especially in minutiae details on mark. Therefore, more work such as using cartridges of other shot numbers to search in the database are needed to verify it.
3.2.3. Ranking positions There are exactly 324 known match cartridges and 1300 known non match cartridges in the database for the first cartridge of each firearm. Attention is paid to find whether all the known match
Table 3 Summary of descriptive statistics for each firearm in each histogram of breech face mark. Specimen under examination
Descriptive statistics Skewness
A0001 B0001 C0001 D0001 E0001
Standard deviation
Mean
A
B
C
D
E
A
B
C
D
E
A
B
C
D
E
0.13 0.66 0.21 0.51 0.34
0.12 0.48 0.15 0.08 0.01
0.11 0.23 2.75 0.16 0.27
0.02 0.11 0.00 0.18 0.16
0.04 0.15 0.31 0.20 0.04
0.04 0.01 0.01 0.01 0.01
0.01 0.04 0.01 0.01 0.01
0.01 0.01 0.03 0.01 0.01
0.01 0.01 0.01 0.05 0.01
0.01 0.01 0.01 0.01 0.06
0.11 0.06 0.06 0.06 0.06
0.06 0.17 0.06 0.06 0.06
0.06 0.06 0.09 0.06 0.06
0.06 0.06 0.06 0.20 0.06
0.07 0.06 0.06 0.06 0.20
86
K. Zhang et al. / Forensic Science International 278 (2017) 78–86
Table 4 Summary of statistics value of match objects for firing pin impression and breech face mark. Pistol
Descriptive statistics Skewness FPI
A B C D E
0.69 0.84 1.10 1.31 1.52
Standard deviation
Mean
BF
FPI
BF
FPI
BF
0.13 0.48 2.75 0.18 0.04
0.06 0.02 0.05 0.05 0.05
0.04 0.04 0.03 0.05 0.06
0.47 0.19 0.36 0.37 0.53
0.11 0.17 0.09 0.20 0.20
Table 5 Number of the match objects ranked within the top 324 positions. Specimen under examination Number of the match objects ranked within the top 324 positions
A0001 B0001 C0001 D0001 E0001
FPI
BF
FPI or BF
324 323 324 324 324
233 319 206 318 309
324 324 324 324 324
objects could be ranked on the top of the correlation list, i.e., within the top 324 positions. In practical application of the Evofinder1 system, firearm examiners usually analyze FPI correlation list and BF correlation list simultaneously, so it will be considered as a “hit” if a match object is placed on the top of correlation list by either FPI or BF. In Table 5, the amount of the match objects ranked within the top 324 positions is given in three columns: FPI correlation result separately, BF correlation result separately, either FPI or BF correlation result. For firing pin impression, all match cartridges can be found within the top 324 positions for pistol A, pistol C, pistol D and pistol E. One cartridge (B2850) is excluded for pistol B. For breech face mark, a range between 5 and 15 out of 324 cartridges (1.5–4.6%) are not included on the top of the list for pistol B, pistol D and pistol E. But for pistol A and pistol C, 91 and 118 out of 324 objects (28.1 and 36.4%) are excluded. Fortunately, although B2850 is out of the top 324 positions in the FPI correlation list, it is ranked within the top 324 positions (the 44th position) in the BF correlation list. Besides, the exclusions in BF correlation list are also included within the top 324 positions in the FPI correlation list. Thus, from the standard of either FPI or BF, 100% of the known match objects can be found and ranked on the top of the correlation list by the Evofinder1 system, as is shown in the last column of Table 5. The reproducibility is further proved by this ranking results, which also remind us a concept of “reference ballistic imaging databases” (RBID), which is aimed at registering all guns in circulation in one big reference database [12–14]. The feasibility of an RBID has been examined by Jan De Kinder et al. in 2003 and revisited by Jan De Ceuster et al. in 2015. To build an effective RBID, it is not enough to find match between two consecutive shots in a big database. In practice, it is important whether the later cartridges fired after extensive shots can be identified as a “hit” with the origin fired cartridge by the system. In this paper, we find that it is possible for the Evofinder1 system to perform such identifications over 3070 shots.
3.3. Limitations The limitation of our results lies in that the current database is constructed of cartridges fired in only five firearms. Furthermore, this research is carried out with one single weapon model of a manufacturer. To strength the conclusion, more cartridges bearing the similar class characteristics, which are fired by many other pistols of the same model and different model, need to be added into the database. At the same time, only the first cartridge of each firearm is performed correlation now, and cartridges with other shot numbers should be used to correlate in the database in further study. 4. Conclusions The reproducibility of marks on cartridges fired from five Chinese Norinco QSZ-92 9 19 mm pistols over the course of 3070 shots is proved using the Evofinder1 system, through both onscreen examination and automatic correlation. With the increasing of the shot number, both firing pin impression and breech face mark change slightly. However, these changes have not prohibited a positive identification made by firearm examiners and a correct correlation associated by Evofinder1 system for the first and later fired cartridges. The 1625 cartridge cases fired in five different firearms can be reliably classified based on the standard of either firing pin impression or breech face mark, when the first cartridge of each firearm is used to perform a correlation. The results of our study are in agreement with the work conducted previously and strengthen their findings statistically. Acknowledgment This work was supported by the People’s Public Security University of China [grant number 2016JKF01102]. References [1] J. Hamby, Identification of projectiles, AFTE J. 6 (5–6) (1974) 22. [2] R. Shem, P. Striupaitis, Comparison of 501 consecutively fired bullets and cartridge cases from a 25 caliber Raven pistol, AFTE J. 15 (3) (1983) 109–112. [3] S. Kirby, Comparison of 900 consecutively fired bullets and cartridge cases from a .455 caliber S&W revolver, AFTE J. 15 (3) (1983) 113–126. [4] Y. Ogihara, M. Kubota, M. Sanada, et al., Comparison of 5000 consecutively fired bullets and cartridge cases from a 45 caliber M1911A1 pistol, AFTE J. 15 (3) (1983) 127–140. [5] M.S. Bonfanti, J. De Kinder, The influence of the use of firearms on their characteristic marks, AFTE J. 31 (1) (1999) 318–323. [6] F. Vinci, C.P. Campobasso, J.A. Bailey, Morphological study of class and individual characteristics produced by firing 2500 cartridges in a .45 caliber semi-automatic pistol, AFTE J. 37 (4) (2005) 368. [7] J. Gouwe, J.E. Hamby, S.A. Norris, Comparison of 10,000 consecutively fired cartridge cases from a model 22 Glock. 40 S&W caliber semiautomatic pistol, AFTE J. 40 (1) (2008) 57. [8] A.Y. Sarıbey, A.G. Hannam, Ç. Tarımcı, An investigation into whether or not the class and individual characteristics of five Turkish manufactured pistols change during extensive firing, J. Forensic Sci. 54 (5) (2009) 1068–1072. [9] D. Mikko, J. Miller, J. Flater, Reproducibility of toolmarks on 20,000 bullets fired through an M240 machine gun barrel, AFTE J. 44 (3) (2012) 248–253. [10] T.L. Grom, W.E. Demuth, IBIS correlation results of cartridge cases collected over the course of 500 firings from a Glock pistol, AFTE J. 44 (4) (2012) 361– 363. [11] http://evofinder.com/. [12] J. De Kinder, Ballistic fingerprinting databases, Sci. Justice 42 (4) (2002) 197– 203. [13] J. De Kinder, F. Tulleners, H. Thiebaut, Reference ballistic imaging database performance, Forensic Sci. Int. 140 (2) (2004) 207–215. [14] J. De Ceuster, S. Dujardin, The reference ballistic imaging database revisited, Forensic Sci. Int. 248 (2015) 82–87.
Contents lists available at ScienceDirect
Forensic Science International journal homepage: www.elsevier.com/locate/forsciint
Reproducibility of characteristic marks on fired cartridge cases from five Chinese Norinco QSZ-92 9 19 mm pistols Kaifeng Zhanga , Yaping Luob,* , Peng Zhouc a b c
School of Forensic Science, People’s Public Security University of China, No.1, Muxidi South Street, Xicheng District, Beijing 100038, China Graduate School, People’s Public Security University of China, No.1, Muxidi South Street, Xicheng District, Beijing 100038, China School of Combat Skills and Police Tactics, People’s Public Security University of China, No.1, Muxidi South Street, Xicheng District, Beijing 100038, China
A R T I C L E I N F O
Article history: Received 28 December 2016 Received in revised form 3 May 2017 Accepted 27 June 2017 Available online 4 July 2017 Keywords: Firearms identification Evofinder Ballistic database Automatic correlation
A B S T R A C T
Reproducibility of characteristic marks on fired cartridge cases from five Chinese Norinco QSZ-92 9 19 mm pistols over the course of 3070 shots is addressed using an Evofinder1 system. The first 20 cartridges are all studied, while one random sample out of every ten consecutively fired cartridge cases is studied for the rest 3050 rounds. As such, a total of 325 cartridges are entered into the system for each pistol and a database consisting of 1625 cartridges from five pistols is established in this paper. Both onscreen examination and automatic correlation are used to examine the reproducibility. In the onscreen examination process, it is possible for firearm examiners to positively identify the first cartridge to the 500th, 1000th, 1500th, 2000th, 2500th, 3000th, 3070th cartridge. In the automatic correlation process, the first cartridge of each firearm is separately performed correlation against the database. It is found that the similarity score for match objects changes slightly as the shot number increases, indicating slight variations of marks. However, these variations have not prohibited the Evofinder1 system to make correct correlation and rank 100% of the known match objects on the top of the correlation list by either firing pin impression or breech face mark. Thus, the reproducibility is proved statistically and objectively in this paper. © 2017 Elsevier B.V. All rights reserved.
1. Introduction Fired cartridge cases are used to identify the specific firearm that fired them, based on firing pin impressions, breech face marks, ejector marks and many other characteristic marks left by the firearm during firing process. Firing pin impressions are indentations created when the firing pin of a firearm strikes the primer of cartridge case. If the nose of the firing pin has manufacturing imperfections or damages, these potentially unique characteristics can be impressed into the metal of the primer by the strike. The breech face leaves its characteristic marks when the head of a cartridge case slams against the breech face due to the pressure developed by the combustion gases within the cartridge. Breech marks can be found as parallel breech marks, circular breech marks or no obvious pattern, just as a stippled or granular appearance. Ejector marks are generated when cartridge cases are ejected from the action of a firearm. All of these above marks can be used for firearms identification.
* Corresponding author. Tel.: +86 10 83903102; fax: +86 10 83903102. E-mail address: [email protected] (Y. Luo). http://dx.doi.org/10.1016/j.forsciint.2017.06.032 0379-0738/© 2017 Elsevier B.V. All rights reserved.
However, these marks may be affected by the change of the characteristics on the firing pin, breech face and ejector owing to the use of the firearm. The influence of the use of firearm on the reproducibility of characteristic marks on cartridge cases has been researched in numerous articles [1–10]. In general, no significant changes have been observed and it is possible for firearm examiners to identify the first and last fired cartridges as to be fired in the same firearm [5]. In previous researches, it is mainly by microscopic comparison that firearm examiners examine cartridges fired in the same firearm, which is an extremely time-consuming process. Therefore, only a few number of samples of cartridges of different shot numbers could be compared. Besides, it strongly depends on firearm examiners’ subjective interpretation whether sufficient agreement exists between the first and later fired cartridges or not. In a paper by Grom and Demuth, an IBIS BrassTRAX-3DTM System was employed to examine the reproducibility [10]. A GlockTM pistol was shot for 500 firings and cartridge numbered 1, 10, 25, 50, 100, 150, 200, 250, 300, 350, 400, 450, 500 were recovered and introduced into the system. The first cartridge was retained in a database which contain other 908 cartridges that bear the similar class characteristics as a known test, and then the later
K. Zhang et al. / Forensic Science International 278 (2017) 78–86
fired cartridges were correlated against the database. Consequently, the first cartridge was successfully found as a “hit” for all twelve cartridge samples, demonstrating that the individual characteristics did not change significantly throughout 500 firings, which made the IBIS1 system could correctly associated them. In addition, a trend of minor drop of firing pin impression rankings was observed while breech face rankings remained consistently near the top as the shot number grew. To survey the reproducibility of characteristic marks on fired cartridge cases from five Chinese Norinco QSZ-92 9 19 mm pistols, a method similar to Grom’s is used in this study. The reproducibility is examined using an Evofinder1 system, which is developed by ScannBI Technology. Both onscreen examination and automatic correlation examination are applied, especially the latter. In the correlation process, a large number of cartridges fired from the same firearm are automatically compared by the Evofinder1 system, where the reproducibility and variation of marks are presented by a similarity score objectively. 2. Materials and methods Firearms used to perform the test are five QSZ-92 9 19 mm pistols designed by Norinco. As shown in Table 1, these firearms will be called A, B, C, D and E, respectively, throughout the paper. The ammunition is DAP92-9 ammunition which is also manufactured by Norinco. A total of 3070 rounds were fired for each firearm and the firearms had not been cleaned during the test. Firearms were fired in a cleared area and the cartridge cases were recovered in the following way. The first 20 cartridge cases fired by the firearms were collected and saved in individual envelopes. From the 21st to the 3070th round, every ten consecutively fired cartridge cases were recovered and placed into one envelope together, thus fire sequence was not differentiated any more among these ten cartridges. Firing date, firearm name and number of shots were recorded on the envelope. To facilitate the project, 3070 cartridge cases fired from each firearm were not all used for further study. The first 20 cartridge cases were all studied. For the rest 3050 cartridge cases, it was chosen to select one random sample out of ten consecutive cartridge cases in the very envelope as mentioned. The selected cartridge cases were studied as the 30th, 40th, . . . , 3070th round. Therefore, 325 cartridge cases were selected for each firearm, and 1625 cartridge cases were retained for study in total. The Evofinder1 system applied in this paper is composed of a data acquisition station (DAS), a specimen analysis system (SAS) with the software version 6.3 and an expert working station (EWS). DAS is aimed at digitally recording objects surfaces and saving them into the database. SAS is aimed at performing an automatic identification of the object under examination against objects in the database and storing high-quality digital images with corresponding information. EWS is aimed at onscreen identification as well as sending a request to SAS for automatic correlation procedure start. EWS provides with hit-lists of objects previously saved in the database and similar to the object under examination [11]. Objects in the database are ranked according to a similarity score ranging from 0 to 1, which is automatically calculated on the Table 1 Five firearms used to perform the test firing. Firearm
Serial number
A B C D E
012160 007252 012375 012624 010457
79
basis of system’s algorithm. Higher the similarity score is, the object in the database looks more similar to the object under examination. The selected 1625 cartridge cases were entered into the Evofinder1 system under its manual protocol. We cleaned, oriented and scanned all of the cartridges in the same way. It was decided to orient the ejector mark at 9 o’clock position. Cartridges were named by a combination of pistol name and shot number in the database. For example, A0001 represents the first cartridge from pistol A, B0500 means one of the 491th–500th cartridge cases from pistol B and so forth. As such, a database of 1625 cartridge cases which could be equally divided into five groups according to the source was established. Afterwards, onscreen examination of cartridge cases was conducted. Images of three-dimensional information of the cartridge markings were compared side by side on the screen. For each firearm, firstly the beginning 20 cartridges were intercompared to build the elements of identity and secondly the 500th, 1000th, 1500th, 2000th, 2500th, 3000th and 3070th cartridges were separately compared to the first cartridge. In the procedure of automatic correlation, the first cartridge of each firearm (A0001, B0001, C0001, D0001 and E0001) was run correlation against the database separately. It should be noted that when one of the above cartridge cases is used for search, the cartridge itself will not be correlated while other 1624 cartridges are all correlated. On starting a correlation, we chose to search by firing pin impression and breech face mark and abandoned ejector mark because of its poor quality. The optional function of “Breech nonoriented”, which is usually used in situation where cartridge is scanned without orientation, was not selected in this study. 3. Results and discussions 3.1. Onscreen examination 3.1.1. Preliminary examination There exist several marks on the bottom of fired cartridge case, including class characteristics and individual characteristics. The class mark displaying an interesting horseshoe shape at the head of the cartridge is impressed by the breech bolt. It is called “horseshoe mark” by Chinese firearm examiners. As to individual mark, the firing pin impression shows ordinary circle shape while the flow back mark surrounding it looks extraordinarily. The breech face mark is found as normal arc and line. The general appearance of a fired cartridge case is illustrated in Fig. 1. 3.1.2. Identity elements By inter-comparing images of the first 20 cartridges, the identity elements of marks for each firearm was established. It is found that five pistols produce similar class characteristics as well as dissimilar individual characteristics. Although the class characteristics are extremely alike, minute differences can still be observed. For example, the horseshoe mark, which is a typical class characteristic, appears almost the same on cartridges from different guns, yet the curvature of the horseshoe shape differs. As to the individual characteristics, either firing pin impression or breech face mark can be used to identify these cartridges to one of the five firearms separately. For firing pin impression, not only the tiny marks lying at the bottom of the aperture but the flow back marks are different as well. Breech face mark also shows distinctive for each firearm. 3.1.3. Reproducibility Comparing the first cartridge to the 500th, 1000th, 1500th, 2000th, 2500th, 3000th, 3070th cartridge, characteristics on the first cartridge can also be observed on the later fired cartridge
80
K. Zhang et al. / Forensic Science International 278 (2017) 78–86
Fig. 1. Image on the left indicates the “horseshoe mark” by a yellow dotted line and the flow back mark by a blue solid line. The same image without annotations is on the right in comparison.
samples, which means characteristics are reproducing. As the shot number increases, slight variation of the characteristics is noted for both firing pin impression and breech face mark. Images in Fig. 2 indicate firing pin impression and breech face mark of the 1st, 1000th, 2000th and 3000th cartridge of each firearm. 3.2. Automatic correlation
study. For breech face mark, the variation of the similarity score conveys interesting information. There may be some special characteristics on breech faces of the five new firearms. These characteristics present irregularly from round to round and disappear completely after tens of firings. We guess that these special characteristics may be caused by some substance existing in breech face of new weapons, thus they are not the real characteristics of the firearm. To check this hypothesis, cartridges of other shot number can be used to perform a correlation in future studies.
When the first cartridge of each firearm is used for search in the database, cartridges with the same source and other four sources are associated equally. To some extent, cartridges fired in other four firearms act as background interference. It should be specified that the horseshoe mark and the flow back mark mentioned in Section 3.1 do not take part in the automatic correlation in the light of system’s algorithm.The correlation results are given in two separate lists: a list of firing pin impression (FPI) and a list of breech face mark (BF).
3.2.2. Distribution of similarity score In this section, the similarity score of both the known match objects and the known non match objects in each correlation list is examined. Using the Minitab1 17 software, the density distribution histograms are charted and the descriptive statistics are calculated.
3.2.1. Similarity score as a function of shot number In theory, if a cartridge is associated with many cartridges fired by the same firearm, the similarity score should be the same, in the case of an ideal reproducibility and the most effective automated ballistic identification system. However, there does not exist an ideal reproducibility, so it is reasonable to get different similarity scores. In this paper, we are interested in verifying if and how the similarity score varies as the shot number increases for each firearm. With the help of filter function, it is easy to exclusively show similarity scores of the known match objects. The similarity score as a function of shot number is examined for firing pin impression and breech face mark separately, as shown in Figs. 3 and 4. In Fig. 3, we can notice a common sign that as the shot number increases, more dots appear at lower units. It demonstrates that the similarity score decreases as the shot number increases. On the contrary, the similarity score for breech face mark is of interest at the very beginning of the shot number, as illustrated in Fig. 4. It is found that the similarity score ranges so rapidly that the dots arrayed almost in a straight line along the y-axis. Towards right along the x-axis, the similarity score keeps in a relatively low level and no regular trend is observed again. The change of the similarity score reveals that firearm signatures evolve over the course of 3070 shots. This parallels the results of onscreen examination. With firing repeatedly, the firing pin of the firearm wear and then the change of firing pin impression on fired cartridges results. Supposing more cartridges are fired in these five guns, a more obvious diminution of the similarity score may be observed, which are not undertaken in this
3.2.2.1. Firing pin impression. Histograms for firing pin impression correlation results are illustrated in Fig. 5. We can learn three interesting things from them. Firstly, the bars representing match objects are built in higher similarity score area while non match objects in lower similarity score area. As can be seen in five histograms, the distance between match and non match objects is notable. This result shows that Evofinder1 has correctly calculated with higher similarity score for known match objects than known non match objects, which indicates a good correlation performance of Evofinder1. Secondly, the distributions of the similarity score for the non match objects are extremely narrow. In contrast, the distribution of the match objects is wide. To explain this situation, we have specially queried personals of Evofinder1 company. In principle, when two cartridges bear more similar areas, the similarity score given by the Evofinder1 system will be higher. Imagine an ideal firearm which exactly reproduces the same imprint of its individual characteristics on every round, and thus any two cartridges fired in this firearm will be associated with similarity score close to 1. Likewise, imagine ideal firearms which own absolutely different individual marks, and then any two cartridges fired in different firearms will be associated with similarity score close to 0. In such perfect scenario, distribution of matches will present high mean value near 1 with low deviation and distribution of non matches will present low mean value near 0 with low deviation too. However, real firearms are not so perfect that neither the reproducibility nor the individuality is absolute, because of which both match distribution and non match distribution are wider.
K. Zhang et al. / Forensic Science International 278 (2017) 78–86
81
Fig. 2. Images are grouped as the following rule: the x-axis represents the shot number and the y-axis represents the letter of weapon.
For match objects, reproduced imprints are sometimes located at the same parts of cartridge as a result of relatively high similarity score (for example 0.8), and sometimes in different parts of cartridge with small intersection area as a result of relatively low
similarity score (for example 0.4 or 0.5). As such, a deviation of the similarity score of match objects can be imagined. For non match objects, though the similarity score will not be equal too, nevertheless, they are usually very low (typically lower than
82
K. Zhang et al. / Forensic Science International 278 (2017) 78–86
Fig. 3. Graphs representing the similarity score as a function of the shot number. The similarity score in A0001 FPI, B0001 FPI, C0001 FPI, D0001 FPI and E0001 FPI stands for the known match objects in the FPI correlation results of A0001, B0001, C0001, D0001, E0001 correspondingly.
0.15), resulting in a smaller deviation in comparison to match objects. Thus, it is logical that the distribution of matches is wide while the distribution of non matches is narrower in this paper. Reproducibility and individuality are correctly presented in this correlation result, which is another indicator of the good correlation performance of Evofinder1. In real cases, a narrow distribution of non match objects is needed to improve the probability of match object to be found. The narrower the distribution of non match objects, the probability is less that some insolent non match object will override the similarity score of match objects and kick out the match objects from the top of the ranking list. Thirdly, in all five histograms, the distribution of match objects shows an interesting tail towards lower similarity scores, which indicate that the data is skew to the left. Furthermore, the width of the match data for each firearm (full width at half maximum or standard deviation) appears approximately equal except for pistol B.
In Table 2, the descriptive statistics: skewness, standard deviation and mean for each firearm in each histogram in Fig. 5 are summarized. These data are meaningful, especially the numbers in bold type which stand for the values of match objects. It is found that the absolute value of numbers in bold is maximal in both the row and the column it locates for each statistics. This result is in consistent with the information in histograms. In addition, the negative value of skewness of for match objects has responded to the “tails” towards left in histograms, disclosing the decline trend of the similarity score, wherein the evolution of firing pin impression during extensive shots is further proved. Five pistols have approximately equal standard deviation values and mean values (numbers in bold) except for pistol B, which gives the lowest value of both statistics. The difference in mean value may be explained by the fact that similarity score is determined by the quality of mark. For instance, it is clear that E0001 bears better quality of firing pin impression than B0001 from Fig. 2, and hence the mean value for pistol E is higher than that for pistol B.
K. Zhang et al. / Forensic Science International 278 (2017) 78–86
83
Fig. 4. Graphs representing the similarity score as a function of the shot number. The similarity score in A0001 BF, B0001 BF, C0001 BF, D0001 BF and E0001 BF stands for the known match objects in the BF correlation results of A0001, B0001, C0001, D0001, E0001 correspondingly.
Furthermore, it seems that the lowest mean value has led to the lowest standard deviation for pistol B, which discloses the correlation between mean and standard deviation to some extent. However, more evidence is required to prove it. 3.2.2.2. Breech face mark. The same histograms are made for breech face mark, as shown in Fig. 6. There is a tiny overlap between the bars of match objects and non match objects in all five histograms. Still, the distribution of non match objects is very narrow as opposed to match objects with a wide distribution. Five firearms show about equal width for the distribution of match objects except for pistol C, the data of which looks a bit narrow. Also, the values of skewness, standard deviation and mean of every firearm in each histogram are summarized, as shown in Table 3. The behavior that the value representative of match objects (numbers in bold) is maximal takes place in three statistics apart
from skewness, the value of which has not showed any regularity. Again, for the values of standard deviation and mean (numbers in bold), five firearms show about equal numbers. It is pistol C that gives the lowest value of both statistics this time. The relatively less outstanding characteristics on breech face mark of pistol C, as shown in Fig. 2, can be used to explain the lowest mean value. What is interesting is the assumption that the lowest mean leads to the lowest standard deviation works here. Nevertheless, we can not draw a certain conclusion and more data are needed to check the correlation between mean and standard deviation. 3.2.2.3. Comparison of descriptive statistics for FPI and BF. To compare the statistics value of match objects for two types of marks: firing pin impression and breech face mark, the numbers in bold in Tables 2 and 3 are grouped in Table 4. In the column of skewness, it is found that the values are negative for firing pin impression as opposed to positive for breech
84
K. Zhang et al. / Forensic Science International 278 (2017) 78–86
Fig. 5. Histograms in A0001 FPI, B0001 FPI, C0001 FPI, D0001 FPI and E0001 FPI representing the density distribution of the similarity score under examination of A0001, B0001, C0001, D0001 and E0001 by firing pin impression correspondingly.
face mark. The variation of breech face mark has not been indicated by skewness as clear as that of firing pin impression. For each firearm, the standard deviation value of two types of marks is approximately equal, which indicates that the extent of the reproducibility of firing pin impression and breech face mark on
the first cartridge is at the same level. With regard to the mean value, firing pin impression presents higher value than breech face mark. It may be caused by the different quality of two types of marks on the first cartridge, as well as the difference in respective algorithms. According to the personals of Evofinder1 company, the
Table 2 Summary of descriptive statistics for each firearm in each histogram of firing pin impression. Specimen under examination
Descriptive statistics Skewness
A0001 B0001 C0001 D0001 E0001
Standard deviation
Mean
A
B
C
D
E
A
B
C
D
E
A
B
C
D
E
0.69 0.06 0.36 0.10 0.11
0.27 0.84 0.51 0.03 0.07
0.10 0.60 1.10 0.09 0.06
0.10 0.37 0.21 1.31 0.09
0.08 0.42 0.16 0.11 1.52
0.06 0.01 0.01 0.01 0.01
0.01 0.02 0.01 0.01 0.01
0.01 0.01 0.05 0.01 0.02
0.01 0.01 0.01 0.05 0.02
0.01 0.01 0.01 0.02 0.05
0.47 0.07 0.09 0.08 0.09
0.08 0.19 0.06 0.08 0.09
0.08 0.07 0.36 0.09 0.09
0.09 0.07 0.09 0.37 0.10
0.09 0.06 0.09 0.10 0.53
K. Zhang et al. / Forensic Science International 278 (2017) 78–86
85
Fig. 6. Histograms in A0001 BF, B0001 BF, C0001 BF, D0001 BF and E0001 BF representing the density distribution of the similarity score under examination of A0001, B0001, C0001, D0001 and E0001 by breech face mark correspondingly.
algorithms for breech face and firing pin impression are different, especially in minutiae details on mark. Therefore, more work such as using cartridges of other shot numbers to search in the database are needed to verify it.
3.2.3. Ranking positions There are exactly 324 known match cartridges and 1300 known non match cartridges in the database for the first cartridge of each firearm. Attention is paid to find whether all the known match
Table 3 Summary of descriptive statistics for each firearm in each histogram of breech face mark. Specimen under examination
Descriptive statistics Skewness
A0001 B0001 C0001 D0001 E0001
Standard deviation
Mean
A
B
C
D
E
A
B
C
D
E
A
B
C
D
E
0.13 0.66 0.21 0.51 0.34
0.12 0.48 0.15 0.08 0.01
0.11 0.23 2.75 0.16 0.27
0.02 0.11 0.00 0.18 0.16
0.04 0.15 0.31 0.20 0.04
0.04 0.01 0.01 0.01 0.01
0.01 0.04 0.01 0.01 0.01
0.01 0.01 0.03 0.01 0.01
0.01 0.01 0.01 0.05 0.01
0.01 0.01 0.01 0.01 0.06
0.11 0.06 0.06 0.06 0.06
0.06 0.17 0.06 0.06 0.06
0.06 0.06 0.09 0.06 0.06
0.06 0.06 0.06 0.20 0.06
0.07 0.06 0.06 0.06 0.20
86
K. Zhang et al. / Forensic Science International 278 (2017) 78–86
Table 4 Summary of statistics value of match objects for firing pin impression and breech face mark. Pistol
Descriptive statistics Skewness FPI
A B C D E
0.69 0.84 1.10 1.31 1.52
Standard deviation
Mean
BF
FPI
BF
FPI
BF
0.13 0.48 2.75 0.18 0.04
0.06 0.02 0.05 0.05 0.05
0.04 0.04 0.03 0.05 0.06
0.47 0.19 0.36 0.37 0.53
0.11 0.17 0.09 0.20 0.20
Table 5 Number of the match objects ranked within the top 324 positions. Specimen under examination Number of the match objects ranked within the top 324 positions
A0001 B0001 C0001 D0001 E0001
FPI
BF
FPI or BF
324 323 324 324 324
233 319 206 318 309
324 324 324 324 324
objects could be ranked on the top of the correlation list, i.e., within the top 324 positions. In practical application of the Evofinder1 system, firearm examiners usually analyze FPI correlation list and BF correlation list simultaneously, so it will be considered as a “hit” if a match object is placed on the top of correlation list by either FPI or BF. In Table 5, the amount of the match objects ranked within the top 324 positions is given in three columns: FPI correlation result separately, BF correlation result separately, either FPI or BF correlation result. For firing pin impression, all match cartridges can be found within the top 324 positions for pistol A, pistol C, pistol D and pistol E. One cartridge (B2850) is excluded for pistol B. For breech face mark, a range between 5 and 15 out of 324 cartridges (1.5–4.6%) are not included on the top of the list for pistol B, pistol D and pistol E. But for pistol A and pistol C, 91 and 118 out of 324 objects (28.1 and 36.4%) are excluded. Fortunately, although B2850 is out of the top 324 positions in the FPI correlation list, it is ranked within the top 324 positions (the 44th position) in the BF correlation list. Besides, the exclusions in BF correlation list are also included within the top 324 positions in the FPI correlation list. Thus, from the standard of either FPI or BF, 100% of the known match objects can be found and ranked on the top of the correlation list by the Evofinder1 system, as is shown in the last column of Table 5. The reproducibility is further proved by this ranking results, which also remind us a concept of “reference ballistic imaging databases” (RBID), which is aimed at registering all guns in circulation in one big reference database [12–14]. The feasibility of an RBID has been examined by Jan De Kinder et al. in 2003 and revisited by Jan De Ceuster et al. in 2015. To build an effective RBID, it is not enough to find match between two consecutive shots in a big database. In practice, it is important whether the later cartridges fired after extensive shots can be identified as a “hit” with the origin fired cartridge by the system. In this paper, we find that it is possible for the Evofinder1 system to perform such identifications over 3070 shots.
3.3. Limitations The limitation of our results lies in that the current database is constructed of cartridges fired in only five firearms. Furthermore, this research is carried out with one single weapon model of a manufacturer. To strength the conclusion, more cartridges bearing the similar class characteristics, which are fired by many other pistols of the same model and different model, need to be added into the database. At the same time, only the first cartridge of each firearm is performed correlation now, and cartridges with other shot numbers should be used to correlate in the database in further study. 4. Conclusions The reproducibility of marks on cartridges fired from five Chinese Norinco QSZ-92 9 19 mm pistols over the course of 3070 shots is proved using the Evofinder1 system, through both onscreen examination and automatic correlation. With the increasing of the shot number, both firing pin impression and breech face mark change slightly. However, these changes have not prohibited a positive identification made by firearm examiners and a correct correlation associated by Evofinder1 system for the first and later fired cartridges. The 1625 cartridge cases fired in five different firearms can be reliably classified based on the standard of either firing pin impression or breech face mark, when the first cartridge of each firearm is used to perform a correlation. The results of our study are in agreement with the work conducted previously and strengthen their findings statistically. Acknowledgment This work was supported by the People’s Public Security University of China [grant number 2016JKF01102]. References [1] J. Hamby, Identification of projectiles, AFTE J. 6 (5–6) (1974) 22. [2] R. Shem, P. Striupaitis, Comparison of 501 consecutively fired bullets and cartridge cases from a 25 caliber Raven pistol, AFTE J. 15 (3) (1983) 109–112. [3] S. Kirby, Comparison of 900 consecutively fired bullets and cartridge cases from a .455 caliber S&W revolver, AFTE J. 15 (3) (1983) 113–126. [4] Y. Ogihara, M. Kubota, M. Sanada, et al., Comparison of 5000 consecutively fired bullets and cartridge cases from a 45 caliber M1911A1 pistol, AFTE J. 15 (3) (1983) 127–140. [5] M.S. Bonfanti, J. De Kinder, The influence of the use of firearms on their characteristic marks, AFTE J. 31 (1) (1999) 318–323. [6] F. Vinci, C.P. Campobasso, J.A. Bailey, Morphological study of class and individual characteristics produced by firing 2500 cartridges in a .45 caliber semi-automatic pistol, AFTE J. 37 (4) (2005) 368. [7] J. Gouwe, J.E. Hamby, S.A. Norris, Comparison of 10,000 consecutively fired cartridge cases from a model 22 Glock. 40 S&W caliber semiautomatic pistol, AFTE J. 40 (1) (2008) 57. [8] A.Y. Sarıbey, A.G. Hannam, Ç. Tarımcı, An investigation into whether or not the class and individual characteristics of five Turkish manufactured pistols change during extensive firing, J. Forensic Sci. 54 (5) (2009) 1068–1072. [9] D. Mikko, J. Miller, J. Flater, Reproducibility of toolmarks on 20,000 bullets fired through an M240 machine gun barrel, AFTE J. 44 (3) (2012) 248–253. [10] T.L. Grom, W.E. Demuth, IBIS correlation results of cartridge cases collected over the course of 500 firings from a Glock pistol, AFTE J. 44 (4) (2012) 361– 363. [11] http://evofinder.com/. [12] J. De Kinder, Ballistic fingerprinting databases, Sci. Justice 42 (4) (2002) 197– 203. [13] J. De Kinder, F. Tulleners, H. Thiebaut, Reference ballistic imaging database performance, Forensic Sci. Int. 140 (2) (2004) 207–215. [14] J. De Ceuster, S. Dujardin, The reference ballistic imaging database revisited, Forensic Sci. Int. 248 (2015) 82–87.