- Email: [email protected]
Development of multiple regression model to estimate the apple’s bruise depth using thermal maps
Postharvest Biology and Technology 116 (2016) 75–79
Contents lists available at ScienceDirect
Postharvest Biology and Technology journal homepage: www.elsevier.com/locate/postharvbio
Development of multiple regression model to estimate the apple’s bruise depth using thermal maps Omid Doosti-Irania , Mahmood Reza Golzariana,* , Mohammad Hossein Aghkhania , Hasan Sadrniaa , Mahboobe Doosti-Iranib a b
Department of Biosystems Engineering, Ferdowsi University of Mashhad, Iran Department of Mathematical Sciences, Isfahan University of Technology, Isfahan, Iran
A R T I C L E I N F O
A B S T R A C T
Article history: Received 4 February 2015 Received in revised form 9 December 2015 Accepted 22 December 2015 Available online 24 January 2016
Thermography is a useful technology for non-contact two-dimensional temperature measurement of the material’s surface that requires nondestructive evaluation. It is also considered as a non-destructive method to determine quality attributes of agricultural products. In this study, the surface temperature of bruised apples was determined using thermal maps and the bruise depth during a set of factorial experiments in the form of a completely randomized design with two factors, namely impact location (bottom, middle, top) and impact energy at three levels (200, 700, 1200 mJ). Then, the relationship between bruise depth and surface temperature was investigated using multiple regression analysis. The results of analysis of variance showed that impact energy and the interaction of impact energy and location of impact region had a significant effect on temperature. The results of the multiple regression model showed that the surface temperature measured from a region can be used to predict the bruise depth in that region. The residual analysis confirmed the prediction adequacy of the proposed model. ã 2016 Published by Elsevier B.V.
Keywords: Thermography Bruise depth Impact energy Impact region Mechanical damage
1. Introduction For most fruits including apples bruising is the most common postharvest mechanical damage (Wilson et al., 1999). Bruising means damages occurred to fruit tissue by an external force that causes physical changes in the texture and chemical changes in the color and flavor. This phenomenon is one of the main reasons that led bruised apples be placed in lower quality grades in inspections (Xing and Baerdemaker, 2005). A large amount of fruits are destroyed or degraded due to bruise and other mechanical damages during harvesting, transporting, storage and packaging. More than 30% of apples are damaged during picking, handling and storage operations (Kupferman, 2006). Based on the US grading standard, existence of the damaged area in 16 mm diameter and 1.6 mm depth will exclude get out the apples from high degreeindicators (USDA, 2002). In a study, the economic depreciation of apples resulting from mechanical damage was estimated in Belgium. They found that the percentage of bruised apples in 2000 and 2001 was equal to 15% and 8%, respectively. They concluded that a reduction of only 10% of the amount of bruised apples resulting from mechanical damage can increase revenues
* Corresponding author. E-mail address: [email protected] (M.R. Golzarian). http://dx.doi.org/10.1016/j.postharvbio.2015.12.024 0925-5214/ ã 2016 Published by Elsevier B.V.
with 892 million dollars in 2000 and 595 million dollars in 2001 (Van Zeebroeck et al., 2003). One of the non-destructive methods that can be used to detect damages in fruit is thermal imaging. Thermography is a useful technology for non-contact two-dimensional temperature measurement of the material’s surface that requires nondestructive evaluation (Kheiralipour et al. 2013). This technology is able to detect the subjects and objects that their surface temperature is different from background (Meola and Carlomagno, 2004). Thermography and using thermal maps are used as non-destructive methods to detect defects that can be implemented in two common modes: passive and active thermography. Passive thermography insists on measuring the heat emitted by natural variations in temperature between normal and bruised tissues. This is while in active thermography an external heat source is used for heating the area of study first before being thermally imaged (Meola and Carlomagno, 2004). Thermal images are capable of providing a thermal map from the product surface. Infrared thermography is evolved as a non-contact means for monitoring conditions to display the surface temperature of objects (Bagavathiappan et al., 2013). Varith et al. (2003) used thermography for detection of bruise in apples when 48 h after dropping for the bruises to develop fully. The results show the 1–2 C difference between normal and bruised tissues, which is due to the difference in thermal penetration capabilities in these
76
O. Doosti-Irani et al. / Postharvest Biology and Technology 116 (2016) 75–79
tissues. Baranowski et al. (2009) used passive thermography to distinguish crushing in three varieties of apple, namely Janagold, Ligol, Gloster. Temperature change patterns of bruised fruit’s surface revealed that the temperature difference between normal and bruised tissues changes between 0.5–1.5 C. This experiment was performed at room temperature of about 25 C. The results also showed that in passive thermography bruised tissue can be detected 48 h after impact. The most temperature difference is observed in Janagold variety and the least temperature difference is observed in Gloster that may be due to differences in their tissue integrity (Baranowski et al., 2009). In another research, thermography method was used to the diagnosis of watercore in apple. Obtained temperature of apples is considered as an appropriate parameter to evaluate the differences in thermal properties between patients and healthy apples. The rate of temperature rise in the initial stages of heating for apples with watercore defects, were significantly lower than healthy apples. The results showed a good correlation between the changes in the temperature and density of the fruit tissue (Baranowski et al., 2008). Zarifneshat et al. (2010) reported that the size of bruising occurred in upper and lower portions of fruit was lower than other parts. The reason for the difference in bruising rate is expressed in differences in the radius of curvature of the fruit in different parts (Zarifneshat et al., 2010). In another research, Siyami et al. (1998) reported that the radius of curvature of an apple has a significant effect on the size and diameter of bruised area in apple, so that by increasing the radius of curvature of the fruit, the diameter of bruising decreases. Van Zeebroeck et al. (2007) evaluated the effective factors on bruise in apple and reported that the apples’ harvest time is effective on bruise. Much research has been done to investigate the important factors on bruising is important today is that to realize the actual contribution of the bruising indicators such as depth and volume of bruise with a non-destructive method. Bruise depth can be considered as an appropriate parameter for grading bruised apples. Thus, the purpose of this study is to determine the relationship between the recorded temperature in depth and surface of bruised tissue using thermal imaging system.
the calyx, the upper zone close to the stem and the middle zone on the equator. Bruising simulation was performed by impact application and in a completely randomized design form in three levels of energy and three locations of apples and each level with five repeats in the form of the factorial experiment. A pendulum (with the flatsurface head with a mass of 0.796 kg and its arm length of 0.196 m) was used to apply impact force on the samples. The impact energy E (J) is determined by: E ¼ mghð1 cosaÞ
ð1Þ
with E the impact energy (J), m the mass of the pendulum (kg), g the acceleration due to gravity (m s2), h the distance from rotation center to center of gravity of the pendulum in meters, a the rotational angle ( ). Based on this relationship, the three energy levels of 200, 700, and 1200 mJ were implemented with rotational angles of 25 , 60.8 and 78.5 , respectively (Sadrnia and Emadi, 2012). In order to avoid damage to other parts of the apples, the apple was covered with wax. After hitting and crushing simulation, samples were kept at a temperature of 5 C for 624 h. 2.2. Thermal imaging Thermal imaging system included a thermal camera (NEC Avio Infrared Technologies InfRec G100Ex, Japan) that was able to image in the temperature range of 40 to 1500 C, with the resolution of 0.08 C and thermal image of 240 320 pixels. The camera’s spectral range was 8–14 mm. Also, the imaging system had a camera bracket, an insulated compartment and a computer system (Fig. 2). The emissivity of apples was determined to be equal to 0.94. Thermal imaging was done in early afternoon at about 2–4 pm. 2.3. Measurement of bruise depth
2. Materials and methods
In order to measure the bruise of each sample, the bruised area in each sample was cut by metal blade and the bruise depth was measured using a digital caliper with an accuracy of 0.01 mm (Fig. 3).
2.1. Simulation of bruise
2.4. Extraction of desired characteristics from thermal images
To perform this test, 45 samples of Golden Delicious apples were prepared with an average weight of 119 g. While preparing samples, the apples were selected. These apples had no prior decay. To simulate the bruise of apple’s tissue, three points were identified on each apple in three areas include the middle, upper and lower zones (Fig. 1), where the lower zone is located close to
After imaging of samples during the specified period in visible and infrared spectral ranges, visible images and thermal maps were analyzed using the software MATLAB version R2011a (Mathworks Inc., US) and InfRec Analyzer NS9500 (NEC Inc., Japan). The thermal maps were transfered to InfRec Analyzer software and the temperatures of all points were saved in the form
Fig. 1. Bruising in three regions on an apple surface: lower (a), middle (b), upper (c) region.
O. Doosti-Irani et al. / Postharvest Biology and Technology 116 (2016) 75–79
Fig. 2. Schematic diagram of the thermal imaging enclosure and its components (top); example thermal images (bottom).
Fig. 3. Presentation of bruise depth in three areas of an apple. Lower (a), middle (b), and upper (c) region.
Fig. 4. Extraction of desired characteristics from thermal images.
77
78
O. Doosti-Irani et al. / Postharvest Biology and Technology 116 (2016) 75–79
Fig. 5. From left to right: images after each processing step for segmentation of bruised area from thermal images of apples.
of a matrix (table). The thermal matrix was imported into MATLAB and converted to a grayscale image for further analysis using Eq. (2). Iði;jÞ ¼
T ði;jÞ T min T max T min
ð2Þ
Iði;jÞ is the new value of the point (i,j) in the obtained gray image, T ði;jÞ is the temperature at the point (i,j) in the thermal matrix, Tmax and Tmin are the maximum and minimum temperatures in the thermal matrix, respectively. After applying the appropriate pre-processing operations on the visible image to eliminate noise, the processed image was converted to a binary image to extract the region of interest. Then, by multiplying the resulted binary image to the thermal grayscale image, the background was removed from the thermal image. Then, pixel values of bruised and healthy tissue were extracted from this segmented image (Fig. 4). As the values of pixels in this image were normalized numbers between zero and one, Eq. (2) was used in order to achieve original values of the thermal matrix T(i,j). Fig. 5 shows the images obtained after each processing step. 3. Results and discussion The research question for this study was whether there is a relationship between physical properties of bruised tissues and their temperature. This was investigated with a completely randomized design with three levels of energy and three impact locations on apple surfaces and each level with five repetitions in the form of a full factorial experiment. First the effect of these factors on temperature was investigated, and in a next step a multiple regression was used to determine the relationship between the bruise depth and temperature and impact location. The statistical analyses were performed in SPSS-ver18 (IBM Inc., US). Before estimating the factorial model, the assumption of equal variance in each of the 9 treatments was examined. Levene’s test showed that the homogeneity of variance was 0.545 and the homogeneity of variances was thus accepted at a significance level of 1%. Analysis of variance was then used to estimate the introduced effects in the model on the temperature. The results showed that the main effects (energy and location) and their interactions had both an effect at a significance level of 0.05 on the temperature of bruised tissues. Also, the adjusted determination coefficient for the model was estimated to 0.415. In other words, 41.5% changes in temperature were described by energy level and location. Considering that the energy level was significant in the model, Duncan’s multiple range test was used to examine the differences between the means of three levels. The results of these tests are summarized in Table 1. According to Duncan’s test, the average temperature at energy levels of 700 and 1200 mJ was not
significantly different, while they were when the impact energy was 200 mJ. As the bruised tissue is converted to cork tissue and moisture is lost over time, the air is replaced in the intercellular space that was filled with the intracellular fluid. The more the impact energy on the tissues of apple, the harder they will become. In other words, the level of damages to cellular tissues will increase as impact force rises and it is expected that the bruised tissue turns porous and corky. The tissue that was bruised under the impact with the energy level of 200 mJ becomes corklike faster than those under impact energy of 700 and 1200 mJ. In the corky tissue (airfilled tissue) the thermal conductivity is lower. In other words, the heat penetration in bruised corky tissue when was impacted by the energy level 200 mJ, was lower than those affected by 700 and 1200 mJ impacts. Therefore, the temperature of bruised tissue at the energy level of 200 is higher than the bruised tissues of the other two energy levels (Table 1). At energy levels of 700 and 1200 mJ in all three locations the apple surface temperature is lower than 200 energy levels. Table 2 gives descriptive statistics about the bruising depths for nine experimental treatments. As can be seen, the maximum average was related to the energy level of 200 mJ and on the upper region of apples with 95% confidence interval. Next, we investigated the relationship between depth of bruising as a dependent variable and surface temperatures (quantitative) on different regions on the apples (categorical), which are set of independent variables. The quantitative variables are entered into a regression model. However, qualitative variables are entered through indicator variables (Draper and Smith, 1998). Our proposed multiple regression model is as follows: Edepth ¼ b0 þ b1 T þ b2 R2 þ b3 R3 In this model, E is the expected value of the response variable, which is the bruise depth. T is temperature, which is the independent variable and a measurable quantity, on different regions of apples, which is a qualitative variable. The region on the sample surfaces where bruising occurred (top, middle and bottom) is incorporated into the model using two indicator variables R2 and R3. R2 = 1 and R3 = 0 indicates the middle part, and R2 = R3 = 0 indicates the top and R2 = 0, R3 = 1 indicates the bottom part on apple surface. It is worth noting that both R2 and R3 cannot get a value of unity at the same time when we predict the bruise depth on the middle part or bottom part of apple separately. In Table 1 Duncan’s test for equality of mean temperature at different levels of energy. Subgroup Energy (J)
Number
1
2
700 1200 200 Significance level
15 15 15 –
10.42 C 10.49 C – 0.82
– – 11.93 C 1.00
O. Doosti-Irani et al. / Postharvest Biology and Technology 116 (2016) 75–79
79
Table 2 Descriptive statistics of the temperature ( C) of upper, middle and lower regions on the surface of apples bruised by three levels of impact energy. Energy (mJ)
Region
N
Mean
Std. deviation
Std. error
95% confidence interval for mean Lower bound
Upper bound
200 200 200 700 700 700 1200 1200 1200 Total
Lower Middle Upper Lower Middle Upper Lower Middle Upper
5 5 5 5 5 5 5 5 5 45
11.5 11.8 12.3 11.4 10.3 9.4 10.1 10.7 10.5 10.9
1.06 0.31 0.77 0.83 1.02 0.62 1.11 1.4 0.94 1.24
0.47 0.14 0.34 0.37 0.45 0.28 0.49 0.62 0.42 0.18
10.2 11.4 11.4 10.4 9.0 8.6 8.8 9.0 9.3 10.5
12.8 12.2 13.3 12.5 11.6 10.1 11.5 12.4 11.7 11.3
Table 3 Estimation of regression coefficients and their significance test. Non-standardized coefficients
Minimum
Maximum
10.6 11.5 11.2 10.5 9.4 8.7 8.4 8.8 9.5 8.4
13.3 12.4 13.3 12.2 11.8 10.4 11.4 12.7 11.7 13.3
Also, the adjusted coefficient of determination was equal to 90.5, indicating the suitability of the model. References
Model
b
Standardized coefficients
Tstatistics
Significance level
Constant Temperature Bottom Middle
1.679 0.137 3.934 1.318
0.72 0.06 0.19 0.19
2.30 2.07 19.80 6.66
0.02 0.04 0.00 0.00
order to be able to predict the depth of bruise, it is necessary to estimate the model parameters and to test the significance of each of these parameters too. Related estimation and tests are shown in Table 3. As can be seen in Table 3, the effects of all variables in the model were significant. The suitability of these models for estimating the depth of bruise was examined by analyzing the regression model residuals and examining the normality of residuals. The significant level of a sample Kolmogorov–Smirnov test to verify the normality of the distribution of standardized residuals was equal to 0.836 that represents the normality of distribution. No trend was found in the residuals plot. 4. Conclusion Given that apple is one of the strategic products of agriculture in Iran, efforts to reduce waste of this fruit cause prosperity and progress in the agricultural industry. Bruising of apples is a fundamental problem in postharvest handling. Considering that bruising eventually leads to fruit loss over time, its detection using non-destructive testing is essential. In this study, a model is presented using thermal maps prepared from the surface of the apple to estimate the depth of bruise. The results of the multiple regression model showed that the temperature of each zone can be used to estimate the bruise in different regions on apple surface.
Bagavathiappan, S., Lahiri, B.B., Saravanan, T., Philip, J., Jayakumar, T., 2013. Infrared thermography for condition monitoring—a review. Infrared Phys. Technol. 60, 35–55. Baranowski, P., Lipecki, J., Mazurek, W., Walczak, R.T., 2008. Detection of watercore in ‘Gloster’ apples using thermography. Postharvest Biol. Technol. 47, 358–366. Baranowski, P., Mazurek, W., Witkowska-Walczak, Barbara, Sławinski, C., 2009. Detection of early apple bruises using pulsed-phase thermography. Postharvest Biol. Technol. 53, 91–100. Draper, N.R., Smith, H., 1998. Applied Regression Analysis, 3rd edition John Wiley & Sons, New York, US. Kheiralipour, K., Ahmadi, H., Rajabipour, A., Rafiee, S., Javan-Nikkhah, M., Jayas, D.S., 2013. Development of a new threshold based classification model for analyzing thermal imaging data to detect fungal infection of pistachio kernel. Agric. Res. 2, 127–131. Kupferman, E., 2006. Minimizing Bruising in Apples Postharvest Information Network. Washington State University, Tree Fruit Research and Extension Center. http://www.goodfruit.com/minimizing-bruising-in-apples/. Meola, C., Carlomagno, G.M., 2004. Recent advances in the use of infrared thermography. Meas. Sci. Technol. 15, 27–58. Sadrnia, H., Emadi, B., 2012. Determine and compare the sensitivity of different varieties of apples to shock loads. Iran. J. Biosyst. Eng. 43, 9–17. Siyami, S., Brown, G.K., Burgess, G.J., Gerrish, J.B., Tennes, B.R., Burton, C.L., Zapp, RH, 1998. Apple impact bruise prediction models. Trans. ASAE 31, 1038–1046. United States Department of Agriculture (USDA), 2002. United States Standards for Grades of Apples, Washington, D.C. Van Zeebroeck, M., Tijskens, E., Van Liedekerke, P., Deli, V., De Baerdemaeker, J., Ramon, H., 2003. Determination of the dynamical behaviour of biological materials during impact using a pendulum device. Sound Vib. 266, 465–480. Van Zeebroeck, M., Van Linden, V., Darius, P., De Ketelaere, R.H., Tijskens, E., 2007. The effect of fruit factors on the bruise susceptibility of apples. Postharvest Biol. Technol. 46, 10–19. Varith, J., Hyde, G.M., Baritelle, A.L., Fellman, J.K., Sattabongkot, T., 2003. Noncontact bruise detection in apples by thermal imaging. Innovative Food Sci. Emerg. Technol. 4, 211–218. Wilson, L.G., M.D., Boyette, E.A., Estes, 1999. Postharvest Handling and Cooling of Fresh Fruits,Vegetables, and Flowers for Small Farms, North Carolina Cooperative Extension Service. Horticulture Information Leaflet 804. Xing, J., Baerdemaker, J.D., 2005. Fresh bruise detection on selected cultivars apples using visible and NIR spectroscopy. Postharvest Biol. Technol. 45, 176–183. Zarifneshat, S., Ghassemzadeh, H.R., Sadeghi, M., Abbaspour-Fard, M.H., Ahmadi, E., Javadi, A., Shervani-Tabar, M.T., 2010. Effect of impact level and fruit properties on golden delicious apple bruising. Am. J. Agric. Biol. Sci. 5 (2), 114–121.
Contents lists available at ScienceDirect
Postharvest Biology and Technology journal homepage: www.elsevier.com/locate/postharvbio
Development of multiple regression model to estimate the apple’s bruise depth using thermal maps Omid Doosti-Irania , Mahmood Reza Golzariana,* , Mohammad Hossein Aghkhania , Hasan Sadrniaa , Mahboobe Doosti-Iranib a b
Department of Biosystems Engineering, Ferdowsi University of Mashhad, Iran Department of Mathematical Sciences, Isfahan University of Technology, Isfahan, Iran
A R T I C L E I N F O
A B S T R A C T
Article history: Received 4 February 2015 Received in revised form 9 December 2015 Accepted 22 December 2015 Available online 24 January 2016
Thermography is a useful technology for non-contact two-dimensional temperature measurement of the material’s surface that requires nondestructive evaluation. It is also considered as a non-destructive method to determine quality attributes of agricultural products. In this study, the surface temperature of bruised apples was determined using thermal maps and the bruise depth during a set of factorial experiments in the form of a completely randomized design with two factors, namely impact location (bottom, middle, top) and impact energy at three levels (200, 700, 1200 mJ). Then, the relationship between bruise depth and surface temperature was investigated using multiple regression analysis. The results of analysis of variance showed that impact energy and the interaction of impact energy and location of impact region had a significant effect on temperature. The results of the multiple regression model showed that the surface temperature measured from a region can be used to predict the bruise depth in that region. The residual analysis confirmed the prediction adequacy of the proposed model. ã 2016 Published by Elsevier B.V.
Keywords: Thermography Bruise depth Impact energy Impact region Mechanical damage
1. Introduction For most fruits including apples bruising is the most common postharvest mechanical damage (Wilson et al., 1999). Bruising means damages occurred to fruit tissue by an external force that causes physical changes in the texture and chemical changes in the color and flavor. This phenomenon is one of the main reasons that led bruised apples be placed in lower quality grades in inspections (Xing and Baerdemaker, 2005). A large amount of fruits are destroyed or degraded due to bruise and other mechanical damages during harvesting, transporting, storage and packaging. More than 30% of apples are damaged during picking, handling and storage operations (Kupferman, 2006). Based on the US grading standard, existence of the damaged area in 16 mm diameter and 1.6 mm depth will exclude get out the apples from high degreeindicators (USDA, 2002). In a study, the economic depreciation of apples resulting from mechanical damage was estimated in Belgium. They found that the percentage of bruised apples in 2000 and 2001 was equal to 15% and 8%, respectively. They concluded that a reduction of only 10% of the amount of bruised apples resulting from mechanical damage can increase revenues
* Corresponding author. E-mail address: [email protected] (M.R. Golzarian). http://dx.doi.org/10.1016/j.postharvbio.2015.12.024 0925-5214/ ã 2016 Published by Elsevier B.V.
with 892 million dollars in 2000 and 595 million dollars in 2001 (Van Zeebroeck et al., 2003). One of the non-destructive methods that can be used to detect damages in fruit is thermal imaging. Thermography is a useful technology for non-contact two-dimensional temperature measurement of the material’s surface that requires nondestructive evaluation (Kheiralipour et al. 2013). This technology is able to detect the subjects and objects that their surface temperature is different from background (Meola and Carlomagno, 2004). Thermography and using thermal maps are used as non-destructive methods to detect defects that can be implemented in two common modes: passive and active thermography. Passive thermography insists on measuring the heat emitted by natural variations in temperature between normal and bruised tissues. This is while in active thermography an external heat source is used for heating the area of study first before being thermally imaged (Meola and Carlomagno, 2004). Thermal images are capable of providing a thermal map from the product surface. Infrared thermography is evolved as a non-contact means for monitoring conditions to display the surface temperature of objects (Bagavathiappan et al., 2013). Varith et al. (2003) used thermography for detection of bruise in apples when 48 h after dropping for the bruises to develop fully. The results show the 1–2 C difference between normal and bruised tissues, which is due to the difference in thermal penetration capabilities in these
76
O. Doosti-Irani et al. / Postharvest Biology and Technology 116 (2016) 75–79
tissues. Baranowski et al. (2009) used passive thermography to distinguish crushing in three varieties of apple, namely Janagold, Ligol, Gloster. Temperature change patterns of bruised fruit’s surface revealed that the temperature difference between normal and bruised tissues changes between 0.5–1.5 C. This experiment was performed at room temperature of about 25 C. The results also showed that in passive thermography bruised tissue can be detected 48 h after impact. The most temperature difference is observed in Janagold variety and the least temperature difference is observed in Gloster that may be due to differences in their tissue integrity (Baranowski et al., 2009). In another research, thermography method was used to the diagnosis of watercore in apple. Obtained temperature of apples is considered as an appropriate parameter to evaluate the differences in thermal properties between patients and healthy apples. The rate of temperature rise in the initial stages of heating for apples with watercore defects, were significantly lower than healthy apples. The results showed a good correlation between the changes in the temperature and density of the fruit tissue (Baranowski et al., 2008). Zarifneshat et al. (2010) reported that the size of bruising occurred in upper and lower portions of fruit was lower than other parts. The reason for the difference in bruising rate is expressed in differences in the radius of curvature of the fruit in different parts (Zarifneshat et al., 2010). In another research, Siyami et al. (1998) reported that the radius of curvature of an apple has a significant effect on the size and diameter of bruised area in apple, so that by increasing the radius of curvature of the fruit, the diameter of bruising decreases. Van Zeebroeck et al. (2007) evaluated the effective factors on bruise in apple and reported that the apples’ harvest time is effective on bruise. Much research has been done to investigate the important factors on bruising is important today is that to realize the actual contribution of the bruising indicators such as depth and volume of bruise with a non-destructive method. Bruise depth can be considered as an appropriate parameter for grading bruised apples. Thus, the purpose of this study is to determine the relationship between the recorded temperature in depth and surface of bruised tissue using thermal imaging system.
the calyx, the upper zone close to the stem and the middle zone on the equator. Bruising simulation was performed by impact application and in a completely randomized design form in three levels of energy and three locations of apples and each level with five repeats in the form of the factorial experiment. A pendulum (with the flatsurface head with a mass of 0.796 kg and its arm length of 0.196 m) was used to apply impact force on the samples. The impact energy E (J) is determined by: E ¼ mghð1 cosaÞ
ð1Þ
with E the impact energy (J), m the mass of the pendulum (kg), g the acceleration due to gravity (m s2), h the distance from rotation center to center of gravity of the pendulum in meters, a the rotational angle ( ). Based on this relationship, the three energy levels of 200, 700, and 1200 mJ were implemented with rotational angles of 25 , 60.8 and 78.5 , respectively (Sadrnia and Emadi, 2012). In order to avoid damage to other parts of the apples, the apple was covered with wax. After hitting and crushing simulation, samples were kept at a temperature of 5 C for 624 h. 2.2. Thermal imaging Thermal imaging system included a thermal camera (NEC Avio Infrared Technologies InfRec G100Ex, Japan) that was able to image in the temperature range of 40 to 1500 C, with the resolution of 0.08 C and thermal image of 240 320 pixels. The camera’s spectral range was 8–14 mm. Also, the imaging system had a camera bracket, an insulated compartment and a computer system (Fig. 2). The emissivity of apples was determined to be equal to 0.94. Thermal imaging was done in early afternoon at about 2–4 pm. 2.3. Measurement of bruise depth
2. Materials and methods
In order to measure the bruise of each sample, the bruised area in each sample was cut by metal blade and the bruise depth was measured using a digital caliper with an accuracy of 0.01 mm (Fig. 3).
2.1. Simulation of bruise
2.4. Extraction of desired characteristics from thermal images
To perform this test, 45 samples of Golden Delicious apples were prepared with an average weight of 119 g. While preparing samples, the apples were selected. These apples had no prior decay. To simulate the bruise of apple’s tissue, three points were identified on each apple in three areas include the middle, upper and lower zones (Fig. 1), where the lower zone is located close to
After imaging of samples during the specified period in visible and infrared spectral ranges, visible images and thermal maps were analyzed using the software MATLAB version R2011a (Mathworks Inc., US) and InfRec Analyzer NS9500 (NEC Inc., Japan). The thermal maps were transfered to InfRec Analyzer software and the temperatures of all points were saved in the form
Fig. 1. Bruising in three regions on an apple surface: lower (a), middle (b), upper (c) region.
O. Doosti-Irani et al. / Postharvest Biology and Technology 116 (2016) 75–79
Fig. 2. Schematic diagram of the thermal imaging enclosure and its components (top); example thermal images (bottom).
Fig. 3. Presentation of bruise depth in three areas of an apple. Lower (a), middle (b), and upper (c) region.
Fig. 4. Extraction of desired characteristics from thermal images.
77
78
O. Doosti-Irani et al. / Postharvest Biology and Technology 116 (2016) 75–79
Fig. 5. From left to right: images after each processing step for segmentation of bruised area from thermal images of apples.
of a matrix (table). The thermal matrix was imported into MATLAB and converted to a grayscale image for further analysis using Eq. (2). Iði;jÞ ¼
T ði;jÞ T min T max T min
ð2Þ
Iði;jÞ is the new value of the point (i,j) in the obtained gray image, T ði;jÞ is the temperature at the point (i,j) in the thermal matrix, Tmax and Tmin are the maximum and minimum temperatures in the thermal matrix, respectively. After applying the appropriate pre-processing operations on the visible image to eliminate noise, the processed image was converted to a binary image to extract the region of interest. Then, by multiplying the resulted binary image to the thermal grayscale image, the background was removed from the thermal image. Then, pixel values of bruised and healthy tissue were extracted from this segmented image (Fig. 4). As the values of pixels in this image were normalized numbers between zero and one, Eq. (2) was used in order to achieve original values of the thermal matrix T(i,j). Fig. 5 shows the images obtained after each processing step. 3. Results and discussion The research question for this study was whether there is a relationship between physical properties of bruised tissues and their temperature. This was investigated with a completely randomized design with three levels of energy and three impact locations on apple surfaces and each level with five repetitions in the form of a full factorial experiment. First the effect of these factors on temperature was investigated, and in a next step a multiple regression was used to determine the relationship between the bruise depth and temperature and impact location. The statistical analyses were performed in SPSS-ver18 (IBM Inc., US). Before estimating the factorial model, the assumption of equal variance in each of the 9 treatments was examined. Levene’s test showed that the homogeneity of variance was 0.545 and the homogeneity of variances was thus accepted at a significance level of 1%. Analysis of variance was then used to estimate the introduced effects in the model on the temperature. The results showed that the main effects (energy and location) and their interactions had both an effect at a significance level of 0.05 on the temperature of bruised tissues. Also, the adjusted determination coefficient for the model was estimated to 0.415. In other words, 41.5% changes in temperature were described by energy level and location. Considering that the energy level was significant in the model, Duncan’s multiple range test was used to examine the differences between the means of three levels. The results of these tests are summarized in Table 1. According to Duncan’s test, the average temperature at energy levels of 700 and 1200 mJ was not
significantly different, while they were when the impact energy was 200 mJ. As the bruised tissue is converted to cork tissue and moisture is lost over time, the air is replaced in the intercellular space that was filled with the intracellular fluid. The more the impact energy on the tissues of apple, the harder they will become. In other words, the level of damages to cellular tissues will increase as impact force rises and it is expected that the bruised tissue turns porous and corky. The tissue that was bruised under the impact with the energy level of 200 mJ becomes corklike faster than those under impact energy of 700 and 1200 mJ. In the corky tissue (airfilled tissue) the thermal conductivity is lower. In other words, the heat penetration in bruised corky tissue when was impacted by the energy level 200 mJ, was lower than those affected by 700 and 1200 mJ impacts. Therefore, the temperature of bruised tissue at the energy level of 200 is higher than the bruised tissues of the other two energy levels (Table 1). At energy levels of 700 and 1200 mJ in all three locations the apple surface temperature is lower than 200 energy levels. Table 2 gives descriptive statistics about the bruising depths for nine experimental treatments. As can be seen, the maximum average was related to the energy level of 200 mJ and on the upper region of apples with 95% confidence interval. Next, we investigated the relationship between depth of bruising as a dependent variable and surface temperatures (quantitative) on different regions on the apples (categorical), which are set of independent variables. The quantitative variables are entered into a regression model. However, qualitative variables are entered through indicator variables (Draper and Smith, 1998). Our proposed multiple regression model is as follows: Edepth ¼ b0 þ b1 T þ b2 R2 þ b3 R3 In this model, E is the expected value of the response variable, which is the bruise depth. T is temperature, which is the independent variable and a measurable quantity, on different regions of apples, which is a qualitative variable. The region on the sample surfaces where bruising occurred (top, middle and bottom) is incorporated into the model using two indicator variables R2 and R3. R2 = 1 and R3 = 0 indicates the middle part, and R2 = R3 = 0 indicates the top and R2 = 0, R3 = 1 indicates the bottom part on apple surface. It is worth noting that both R2 and R3 cannot get a value of unity at the same time when we predict the bruise depth on the middle part or bottom part of apple separately. In Table 1 Duncan’s test for equality of mean temperature at different levels of energy. Subgroup Energy (J)
Number
1
2
700 1200 200 Significance level
15 15 15 –
10.42 C 10.49 C – 0.82
– – 11.93 C 1.00
O. Doosti-Irani et al. / Postharvest Biology and Technology 116 (2016) 75–79
79
Table 2 Descriptive statistics of the temperature ( C) of upper, middle and lower regions on the surface of apples bruised by three levels of impact energy. Energy (mJ)
Region
N
Mean
Std. deviation
Std. error
95% confidence interval for mean Lower bound
Upper bound
200 200 200 700 700 700 1200 1200 1200 Total
Lower Middle Upper Lower Middle Upper Lower Middle Upper
5 5 5 5 5 5 5 5 5 45
11.5 11.8 12.3 11.4 10.3 9.4 10.1 10.7 10.5 10.9
1.06 0.31 0.77 0.83 1.02 0.62 1.11 1.4 0.94 1.24
0.47 0.14 0.34 0.37 0.45 0.28 0.49 0.62 0.42 0.18
10.2 11.4 11.4 10.4 9.0 8.6 8.8 9.0 9.3 10.5
12.8 12.2 13.3 12.5 11.6 10.1 11.5 12.4 11.7 11.3
Table 3 Estimation of regression coefficients and their significance test. Non-standardized coefficients
Minimum
Maximum
10.6 11.5 11.2 10.5 9.4 8.7 8.4 8.8 9.5 8.4
13.3 12.4 13.3 12.2 11.8 10.4 11.4 12.7 11.7 13.3
Also, the adjusted coefficient of determination was equal to 90.5, indicating the suitability of the model. References
Model
b
Standardized coefficients
Tstatistics
Significance level
Constant Temperature Bottom Middle
1.679 0.137 3.934 1.318
0.72 0.06 0.19 0.19
2.30 2.07 19.80 6.66
0.02 0.04 0.00 0.00
order to be able to predict the depth of bruise, it is necessary to estimate the model parameters and to test the significance of each of these parameters too. Related estimation and tests are shown in Table 3. As can be seen in Table 3, the effects of all variables in the model were significant. The suitability of these models for estimating the depth of bruise was examined by analyzing the regression model residuals and examining the normality of residuals. The significant level of a sample Kolmogorov–Smirnov test to verify the normality of the distribution of standardized residuals was equal to 0.836 that represents the normality of distribution. No trend was found in the residuals plot. 4. Conclusion Given that apple is one of the strategic products of agriculture in Iran, efforts to reduce waste of this fruit cause prosperity and progress in the agricultural industry. Bruising of apples is a fundamental problem in postharvest handling. Considering that bruising eventually leads to fruit loss over time, its detection using non-destructive testing is essential. In this study, a model is presented using thermal maps prepared from the surface of the apple to estimate the depth of bruise. The results of the multiple regression model showed that the temperature of each zone can be used to estimate the bruise in different regions on apple surface.
Bagavathiappan, S., Lahiri, B.B., Saravanan, T., Philip, J., Jayakumar, T., 2013. Infrared thermography for condition monitoring—a review. Infrared Phys. Technol. 60, 35–55. Baranowski, P., Lipecki, J., Mazurek, W., Walczak, R.T., 2008. Detection of watercore in ‘Gloster’ apples using thermography. Postharvest Biol. Technol. 47, 358–366. Baranowski, P., Mazurek, W., Witkowska-Walczak, Barbara, Sławinski, C., 2009. Detection of early apple bruises using pulsed-phase thermography. Postharvest Biol. Technol. 53, 91–100. Draper, N.R., Smith, H., 1998. Applied Regression Analysis, 3rd edition John Wiley & Sons, New York, US. Kheiralipour, K., Ahmadi, H., Rajabipour, A., Rafiee, S., Javan-Nikkhah, M., Jayas, D.S., 2013. Development of a new threshold based classification model for analyzing thermal imaging data to detect fungal infection of pistachio kernel. Agric. Res. 2, 127–131. Kupferman, E., 2006. Minimizing Bruising in Apples Postharvest Information Network. Washington State University, Tree Fruit Research and Extension Center. http://www.goodfruit.com/minimizing-bruising-in-apples/. Meola, C., Carlomagno, G.M., 2004. Recent advances in the use of infrared thermography. Meas. Sci. Technol. 15, 27–58. Sadrnia, H., Emadi, B., 2012. Determine and compare the sensitivity of different varieties of apples to shock loads. Iran. J. Biosyst. Eng. 43, 9–17. Siyami, S., Brown, G.K., Burgess, G.J., Gerrish, J.B., Tennes, B.R., Burton, C.L., Zapp, RH, 1998. Apple impact bruise prediction models. Trans. ASAE 31, 1038–1046. United States Department of Agriculture (USDA), 2002. United States Standards for Grades of Apples, Washington, D.C. Van Zeebroeck, M., Tijskens, E., Van Liedekerke, P., Deli, V., De Baerdemaeker, J., Ramon, H., 2003. Determination of the dynamical behaviour of biological materials during impact using a pendulum device. Sound Vib. 266, 465–480. Van Zeebroeck, M., Van Linden, V., Darius, P., De Ketelaere, R.H., Tijskens, E., 2007. The effect of fruit factors on the bruise susceptibility of apples. Postharvest Biol. Technol. 46, 10–19. Varith, J., Hyde, G.M., Baritelle, A.L., Fellman, J.K., Sattabongkot, T., 2003. Noncontact bruise detection in apples by thermal imaging. Innovative Food Sci. Emerg. Technol. 4, 211–218. Wilson, L.G., M.D., Boyette, E.A., Estes, 1999. Postharvest Handling and Cooling of Fresh Fruits,Vegetables, and Flowers for Small Farms, North Carolina Cooperative Extension Service. Horticulture Information Leaflet 804. Xing, J., Baerdemaker, J.D., 2005. Fresh bruise detection on selected cultivars apples using visible and NIR spectroscopy. Postharvest Biol. Technol. 45, 176–183. Zarifneshat, S., Ghassemzadeh, H.R., Sadeghi, M., Abbaspour-Fard, M.H., Ahmadi, E., Javadi, A., Shervani-Tabar, M.T., 2010. Effect of impact level and fruit properties on golden delicious apple bruising. Am. J. Agric. Biol. Sci. 5 (2), 114–121.