- Email: [email protected]
Automatic control and adjustment of TV picture geometry
Copyright © IFAC Intelligent Manufacturing Systems. Gramado - RS . Brazil. 1998
AUTOMATIC CONTROL AND ADJUSTMENT OF TV PICTURE GEOMETRY
Joze Mohorko, Peter Planinsic, Zarko Cucej
University of Maribor, Faculty of Electrical Engineering, Computer Science and Information Technology, Laboratory for signal processing and remote control 2000 Maribor, Smetanova 17, Slovenia [email protected]
Abstract: The foundation of computer vision system for testing, controlling and adjusting of the TV sets is a mathematical model of distortion of the TV picture. Adequate accurate model description parameters are given by polynomial of order eight and knowledge based nonlinear, cross coupled relationships between command variables for adjusting of the TV sets and coefficients of polynomial. Copyright @ 1998 IFAC Keywords: Television, Quality control, Polynomials, Model based control, Image recognition, Adaptive, Bad data identification.
1. INTRODUCTION
automatic adjustment with computer added control is essential for high quality and productivity of TV sets production. Beside this, the throughput of the automated adjustment is up to five times better than the expert manual work throughput. The automatic control and adjusting of picture geometry can be perform in two ways:
A strong competition on intemational market forces the entertainment electronic manufacturers into the sharp battle for minimisation of the producing costs and improvement of the products quality. There are many ways to reach these goals. One of them is the movement of production process into the countries with low taxes and cheap labour costs, but more promising and best long-term solution is modernisation and automation of the production process. This solution enables the high productivity and constant quality, which are the most important conditions for the survival on the market.
• •
by using robots or specialised manipulators for replacing the hands of workers by designing the TV sets with digital controlled video processors that allow communication with control system and downloading the correct parameters.
In the article the automatic control and adjustment system is described (Figure 1), which was developed for digitally controlled TV sets based on Philips video processor TDA 8366. The system is installed as a part of production line with capacity of 100 TV sets per hour and is capable to control and to adjust one TV set in less than 30 seconds.
In many productions of the TV sets the adjustment of geometry of picture and acoustic parameters are still performed manually. This is very complex task and requests good skilled worker. Regardless to precision defmed procedure of adjusting, the quality of adjusting is very subjective depended. Therefore the
189
algorithms for horizontal image geometry parameters is the polynomial approximation. The polynomial approximate coefficients are composed with some image parameters determined by recommendations (CCIR, 1974; IEC, 1997). These coefficients are used as measured values in controllers of image properties.
ial~ I
i
II
I
Fig. I. Principle of geometry control and adjustment procedure.
-ce I I Cl) 00
I..,
.~
I;
LJ
2. DESCRIPTION OF THE PROBLEM
For development of automatic geometry adjusting procedure, properties of TV sets should be known. There are existing recommendations (CCIR, 1974; IEC, 1997) which determine the following image parameters: S-correction (SC), vertical amplitude (V A), vertical slope (VS), vertical shift (VSH), horizontal shift (HS), E-W width (EW), E-W parabola width (PW), E-W corner parabola (CP) and E-W trapezium (TC). Unfortunately these parameters are interdependent. Specially the group PW, EW, TC and PC, and group VA, VS, VSH and SC are highly coupled. Therefore dependence between control variables and geometric parameters of TV picture was measured. For measurement of inter-dependence of the parameters, the lattice of lines (Figure 2) has been used.
Reference vaJ ues
Fig. 3. Block scheme of compensation of the picture distortion and control system
3. 1 Model of the TV picture geometry
In ideal case a deflection of the electron-beam in cathode tube is linear depended from the deflection coils currents, e.g. in the vertical y-direction from the saw-tooth shaped currents f vert (50 Hz) and in the
y
horizontal x-direction
f ew
(15625 Hz):
y{t) = kyl (t) x{t) = kJ ew (t) verI
(1)
! : X r' .
Because the deflection system is not ideal and the screen of CTR is flat, the deflection of electron-beam should not be linear dependent to deflection currents. At known non-linearity the compensation is performed by compensation circuits with inverse characteristic of non-linearity (Figure 3). The compensation circuits are controlled by nine control registers (SC, VA, VS, VSH, HS, EW, PW, CP and TC). Control variables are integer numbers from interval (0,63) where number 31 has no influence on geometry properties of picture. Model it is described by parameters for horizontal and vertical control variables. In general, they are described by:
Fig. 2. Test picture. Encountered is negative parabolic distortion.
3. ADJUSTMENT OF TV PICTURE GEOMETRY
The foundation for designing automatic adjustment
190
n-31 a=k - where i
32
I
IJ
E
(0,4) , j n
E
E
b=k m-31 j m 32
in: level of screen brightness, picture contrast, differences in brightness between centre and border of image, reflections, noise and unsharpened image. This is reached by developing of histogram based mechanism for electronic shutter controlling in image acquisition algorithm and by using of correlation based methods (Levine, 1985; Heijden, 1994; Jain, et aI., 1995; Pratt, 1978) for line segments searching algorithms.
(2)
(0,3) ,
(HS, EW, TC, PW, CP)
and m E (VSH, VA, VS, SC).
The horizontal distortion of the test picture (Figure 2) is measured by coefficients of interpolation function, which graph has the same shape as a outer vertical lines of the test picture. On base of intensive tests with different interpolation functions is confmned that polynomial function with satisfactory accuracy is reduced to the polynomial function of order 8:
Weighting factors k" and km (Table I; Table 2) are estimated from characteristics of video processor TDA 8366 1 (PHILIPS, 1993). Table I : Weighting factors k n
0.03
0.1
0.04
0.12
(7)
0.22
where x and y are horizontal and vertical coordinate of the picture. The coefficients of polynomial function are determined by mean squared error method. Coefficient Ko is a criteria for EW, KI for TC, K 2
Table 2: Weighting factors km kVSH
kvs
ksc
0.04
0.14
0.12
for PW and K) for Pc.
The corrected vertical signal is than determined by:
y = y+lly
Vertical geometry parameters are estimated from horizontal lines of lattice.
(3)
where 3.3 Control algorithm lly = -(bo + 0,5b2 ) + (b l + 0,5b2
-
3,103Ib))y +
+ 4,81Ob)y J -1 ,722b5/
By designs of controller the following directives have been given:
(4)
and corrected horizontal signal is
• • (5)
where
• box = a o + alx +
-(a l -I)(a2y +aS 2 - a4y 8)X
(6)
•
accuracy controller should be robust, e.g. (i) to variations of parameters of the controlled object, and (ii) to amount of noise and external disturbances convergence (test must be performed in <30 seconds, therefore only a few iterations are allowed) parallel execution of algorithms.
Considering above directives, general algorithm for adjusting of TV picture geometry has been developed. Control algorithm is based on adaptive I controller with self tuning gain in determined dynamic range (Figure 4). Controller also includes decision logic which carries about:
3.2 Measurement The tested TV set picture is captured by high resolution remote controlled digital TV camera Kodak Megaplus 1.4i with resolution 1317 x 1035 pixels. The main request to the measurement is to be robust
• •
I The TDA 8366 is I2C-bus controlled PALINTSC TV video processor. The deflection control circuit provides a drive pulse for horizontal output stage, a differential saw-tooth current for the vertical output stage and East-West drive current for the EastWest output stage. These signals can be manipulated for geometry correction of the picture (PHI LIPS. 1993).
•
19 1
conditions for end of adjustment procedure identification of unstable adjustment with setting the condition to stop the adjustment and indicating the fail temporary disconnection of inputs, when their reliability is low or if the fatal error is detected
With appropriate grouping of controllers and determining of their sequence the coupling of the controlled parameters becomes a smaller obstacle. Since the robustness of controllers and adjustment procedure is very dependent on valuation of the rough measured data, the big effort has been given to design of algorithm for valuation of measured data, to algorithm for estimation of missing data and to decision algorithm for choosing the correct data.
6. CONCLUSION
Major innovations in this work are newly developed robust methods for polynomial based image recognition, nonlinear mathematical model of circuits for geometry compensation and adaptive control system for nonlinear static object with 9 coupled inputs and 9 outputs. The developed test system is used in production since last year. Achieved results encouraged the user management to redesign all of TV sets, so that they can be automatic tested.
Bad data identification
Meas. value
!T- ' I -~
I
!
Adaptation :...... 1 ( - - - --
Fig. 4. Adaptive I controller with self tuning gain in determined dynamic range The control algorithm is performed on IBM compatible Pentium personal computer (133 MHz, 32Mb RAM). For time demanding tasks the program routine was written in C language (Steams, et aI., 1988). The standard function blocks from Lab VIEW library are used for other functions.
4. USER INTERFACE
Fig. 5. Picture of system for Automatic control and adjustment of TV picture geometry
The user interface was developed with LabVIEW program, running on Windows NT platform. Virtual instruments (VI) for handling control and adjustment system are created. VI supervised complete control and adjustment procedure as well as they allow to run test and service mode in which the particular segment of testing system is checked or particular features of TV sets are controlled and/or adjusted. VI also generate report list with data of TV sets and settled parameters and save as day log file in management database system.
REFERENCES CCIR (1974). CClR recommendation, Radiocommunications. Characteristics of TV systems, publication 624 IEC (1997). lEC recommendation, Radiocommunications. TV receivers. Methods of measurements. Geometrical properties of the picture, publication 107-1 part III PHILIPS (1993). Device specifications, TDA8366 Monolythic Integrated 12C bus controlled PALINTSC TV - Processor Heijden, F. (1994), Image Based Measurement Systems Object resognition and parameter estimation, John Wiley & Sons Ltd., Chicester Levine, M. D. (1985). Vision in man and machine, McGraw-Hill Book Company Jain, R., R. Kasutri, B. G. Schunk (1995). Machine vision, McGraw-Hill, Inc. Steams, Samuel D. (1988). Signal Processing Algorithms, Pretence-Hall, Inc.
5. INCORPORATION IN PRODUCTION LINE
The developed test system occupies two work places in production line. The first work place is used for preparation of TV sets to testing, in the second one, which is in the protection cabin, performs the control and adjustment (Figure 5). The test system communicate with production line automatic by digital signals NEXT, STOP, OK and FAIL.
192
AUTOMATIC CONTROL AND ADJUSTMENT OF TV PICTURE GEOMETRY
Joze Mohorko, Peter Planinsic, Zarko Cucej
University of Maribor, Faculty of Electrical Engineering, Computer Science and Information Technology, Laboratory for signal processing and remote control 2000 Maribor, Smetanova 17, Slovenia [email protected]
Abstract: The foundation of computer vision system for testing, controlling and adjusting of the TV sets is a mathematical model of distortion of the TV picture. Adequate accurate model description parameters are given by polynomial of order eight and knowledge based nonlinear, cross coupled relationships between command variables for adjusting of the TV sets and coefficients of polynomial. Copyright @ 1998 IFAC Keywords: Television, Quality control, Polynomials, Model based control, Image recognition, Adaptive, Bad data identification.
1. INTRODUCTION
automatic adjustment with computer added control is essential for high quality and productivity of TV sets production. Beside this, the throughput of the automated adjustment is up to five times better than the expert manual work throughput. The automatic control and adjusting of picture geometry can be perform in two ways:
A strong competition on intemational market forces the entertainment electronic manufacturers into the sharp battle for minimisation of the producing costs and improvement of the products quality. There are many ways to reach these goals. One of them is the movement of production process into the countries with low taxes and cheap labour costs, but more promising and best long-term solution is modernisation and automation of the production process. This solution enables the high productivity and constant quality, which are the most important conditions for the survival on the market.
• •
by using robots or specialised manipulators for replacing the hands of workers by designing the TV sets with digital controlled video processors that allow communication with control system and downloading the correct parameters.
In the article the automatic control and adjustment system is described (Figure 1), which was developed for digitally controlled TV sets based on Philips video processor TDA 8366. The system is installed as a part of production line with capacity of 100 TV sets per hour and is capable to control and to adjust one TV set in less than 30 seconds.
In many productions of the TV sets the adjustment of geometry of picture and acoustic parameters are still performed manually. This is very complex task and requests good skilled worker. Regardless to precision defmed procedure of adjusting, the quality of adjusting is very subjective depended. Therefore the
189
algorithms for horizontal image geometry parameters is the polynomial approximation. The polynomial approximate coefficients are composed with some image parameters determined by recommendations (CCIR, 1974; IEC, 1997). These coefficients are used as measured values in controllers of image properties.
ial~ I
i
II
I
Fig. I. Principle of geometry control and adjustment procedure.
-ce I I Cl) 00
I..,
.~
I;
LJ
2. DESCRIPTION OF THE PROBLEM
For development of automatic geometry adjusting procedure, properties of TV sets should be known. There are existing recommendations (CCIR, 1974; IEC, 1997) which determine the following image parameters: S-correction (SC), vertical amplitude (V A), vertical slope (VS), vertical shift (VSH), horizontal shift (HS), E-W width (EW), E-W parabola width (PW), E-W corner parabola (CP) and E-W trapezium (TC). Unfortunately these parameters are interdependent. Specially the group PW, EW, TC and PC, and group VA, VS, VSH and SC are highly coupled. Therefore dependence between control variables and geometric parameters of TV picture was measured. For measurement of inter-dependence of the parameters, the lattice of lines (Figure 2) has been used.
Reference vaJ ues
Fig. 3. Block scheme of compensation of the picture distortion and control system
3. 1 Model of the TV picture geometry
In ideal case a deflection of the electron-beam in cathode tube is linear depended from the deflection coils currents, e.g. in the vertical y-direction from the saw-tooth shaped currents f vert (50 Hz) and in the
y
horizontal x-direction
f ew
(15625 Hz):
y{t) = kyl (t) x{t) = kJ ew (t) verI
(1)
! : X r' .
Because the deflection system is not ideal and the screen of CTR is flat, the deflection of electron-beam should not be linear dependent to deflection currents. At known non-linearity the compensation is performed by compensation circuits with inverse characteristic of non-linearity (Figure 3). The compensation circuits are controlled by nine control registers (SC, VA, VS, VSH, HS, EW, PW, CP and TC). Control variables are integer numbers from interval (0,63) where number 31 has no influence on geometry properties of picture. Model it is described by parameters for horizontal and vertical control variables. In general, they are described by:
Fig. 2. Test picture. Encountered is negative parabolic distortion.
3. ADJUSTMENT OF TV PICTURE GEOMETRY
The foundation for designing automatic adjustment
190
n-31 a=k - where i
32
I
IJ
E
(0,4) , j n
E
E
b=k m-31 j m 32
in: level of screen brightness, picture contrast, differences in brightness between centre and border of image, reflections, noise and unsharpened image. This is reached by developing of histogram based mechanism for electronic shutter controlling in image acquisition algorithm and by using of correlation based methods (Levine, 1985; Heijden, 1994; Jain, et aI., 1995; Pratt, 1978) for line segments searching algorithms.
(2)
(0,3) ,
(HS, EW, TC, PW, CP)
and m E (VSH, VA, VS, SC).
The horizontal distortion of the test picture (Figure 2) is measured by coefficients of interpolation function, which graph has the same shape as a outer vertical lines of the test picture. On base of intensive tests with different interpolation functions is confmned that polynomial function with satisfactory accuracy is reduced to the polynomial function of order 8:
Weighting factors k" and km (Table I; Table 2) are estimated from characteristics of video processor TDA 8366 1 (PHILIPS, 1993). Table I : Weighting factors k n
0.03
0.1
0.04
0.12
(7)
0.22
where x and y are horizontal and vertical coordinate of the picture. The coefficients of polynomial function are determined by mean squared error method. Coefficient Ko is a criteria for EW, KI for TC, K 2
Table 2: Weighting factors km kVSH
kvs
ksc
0.04
0.14
0.12
for PW and K) for Pc.
The corrected vertical signal is than determined by:
y = y+lly
Vertical geometry parameters are estimated from horizontal lines of lattice.
(3)
where 3.3 Control algorithm lly = -(bo + 0,5b2 ) + (b l + 0,5b2
-
3,103Ib))y +
+ 4,81Ob)y J -1 ,722b5/
By designs of controller the following directives have been given:
(4)
and corrected horizontal signal is
• • (5)
where
• box = a o + alx +
-(a l -I)(a2y +aS 2 - a4y 8)X
(6)
•
accuracy controller should be robust, e.g. (i) to variations of parameters of the controlled object, and (ii) to amount of noise and external disturbances convergence (test must be performed in <30 seconds, therefore only a few iterations are allowed) parallel execution of algorithms.
Considering above directives, general algorithm for adjusting of TV picture geometry has been developed. Control algorithm is based on adaptive I controller with self tuning gain in determined dynamic range (Figure 4). Controller also includes decision logic which carries about:
3.2 Measurement The tested TV set picture is captured by high resolution remote controlled digital TV camera Kodak Megaplus 1.4i with resolution 1317 x 1035 pixels. The main request to the measurement is to be robust
• •
I The TDA 8366 is I2C-bus controlled PALINTSC TV video processor. The deflection control circuit provides a drive pulse for horizontal output stage, a differential saw-tooth current for the vertical output stage and East-West drive current for the EastWest output stage. These signals can be manipulated for geometry correction of the picture (PHI LIPS. 1993).
•
19 1
conditions for end of adjustment procedure identification of unstable adjustment with setting the condition to stop the adjustment and indicating the fail temporary disconnection of inputs, when their reliability is low or if the fatal error is detected
With appropriate grouping of controllers and determining of their sequence the coupling of the controlled parameters becomes a smaller obstacle. Since the robustness of controllers and adjustment procedure is very dependent on valuation of the rough measured data, the big effort has been given to design of algorithm for valuation of measured data, to algorithm for estimation of missing data and to decision algorithm for choosing the correct data.
6. CONCLUSION
Major innovations in this work are newly developed robust methods for polynomial based image recognition, nonlinear mathematical model of circuits for geometry compensation and adaptive control system for nonlinear static object with 9 coupled inputs and 9 outputs. The developed test system is used in production since last year. Achieved results encouraged the user management to redesign all of TV sets, so that they can be automatic tested.
Bad data identification
Meas. value
!T- ' I -~
I
!
Adaptation :...... 1 ( - - - --
Fig. 4. Adaptive I controller with self tuning gain in determined dynamic range The control algorithm is performed on IBM compatible Pentium personal computer (133 MHz, 32Mb RAM). For time demanding tasks the program routine was written in C language (Steams, et aI., 1988). The standard function blocks from Lab VIEW library are used for other functions.
4. USER INTERFACE
Fig. 5. Picture of system for Automatic control and adjustment of TV picture geometry
The user interface was developed with LabVIEW program, running on Windows NT platform. Virtual instruments (VI) for handling control and adjustment system are created. VI supervised complete control and adjustment procedure as well as they allow to run test and service mode in which the particular segment of testing system is checked or particular features of TV sets are controlled and/or adjusted. VI also generate report list with data of TV sets and settled parameters and save as day log file in management database system.
REFERENCES CCIR (1974). CClR recommendation, Radiocommunications. Characteristics of TV systems, publication 624 IEC (1997). lEC recommendation, Radiocommunications. TV receivers. Methods of measurements. Geometrical properties of the picture, publication 107-1 part III PHILIPS (1993). Device specifications, TDA8366 Monolythic Integrated 12C bus controlled PALINTSC TV - Processor Heijden, F. (1994), Image Based Measurement Systems Object resognition and parameter estimation, John Wiley & Sons Ltd., Chicester Levine, M. D. (1985). Vision in man and machine, McGraw-Hill Book Company Jain, R., R. Kasutri, B. G. Schunk (1995). Machine vision, McGraw-Hill, Inc. Steams, Samuel D. (1988). Signal Processing Algorithms, Pretence-Hall, Inc.
5. INCORPORATION IN PRODUCTION LINE
The developed test system occupies two work places in production line. The first work place is used for preparation of TV sets to testing, in the second one, which is in the protection cabin, performs the control and adjustment (Figure 5). The test system communicate with production line automatic by digital signals NEXT, STOP, OK and FAIL.
192