Light source optimization method for supplementary lighting system of visual inspection system and experimental device used

Abstract

The invention discloses a light source optimization method for a fill light system of a visual detection system and an experimental device used therein. The method comprises the following steps: designing an experimental device, comprising a fill light source bracket fixed on an experimental table, an LED matrix arranged on the fill light bracket, an Arduino control board for controlling LEDs, an illuminometer for detecting brightness, an external light source for simulating actual detection ambient light, a light source controller and a processor for calculation; then calculating the spatial posture parameters of the fill light source, establishing a fill light source mathematical model to obtain an illumination vector group arranged in parallel, simplifying the fill light system by using singular value decomposition and establishing a plane illumination effect evaluation function, obtaining an LED illumination weight with an optimal illumination effect, and adjusting the brightness of the fill light source so that a target area is evenly illuminated. The invention simplifies the fill light system model on the basis of ensuring the illumination effect of the visual detection system, and reduces the requirements for an LED hardware control circuit and a brightness weight optimization algorithm.

Description

Technical Field

[0001] The present invention relates to the technical field of industrial visual inspection application, and in particular to a method for optimizing the light source of a fill light system of a visual inspection system and an experimental device used therein. Background Art

[0002] During the visual inspection of integrated circuit chip packaging quality, tray loading or vibrating hopper loading is usually adopted according to different customer needs. The tray loading method uses a method of simultaneously inspecting multiple chips, which has higher inspection efficiency, but also puts forward higher requirements on the lighting effect of visual inspection. Existing single chip visual defect detection (such as patent application number 201810869324.7) or image acquisition of multiple chips (such as patent application number 202011628232.3) are mainly based on optimization algorithms, and the impact of lighting effects on visual inspection systems is not fully considered. The research on existing light source optimization methods is mainly concentrated in darkroom environments, and the impact of the surrounding environment on lighting effects in actual inspections is not fully considered, such as patent application number 202010560546.8 and patent application number 202010513967.5. Therefore, it is necessary to make further improvements to the light source optimization results in a darkroom environment. In terms of the light source optimization method of the fill light system, the paper "Adaptive Brightness Control System of LED Light Source for Visual Inspection" (Office Automation, 2020, 25(01): 44-45+17) and the patent with application number 202010206895.X both use photosensitive sensors or visual sensors to sense the external light intensity, and the microcontroller outputs PWM pulses to control the LED brightness, so that the system dynamically adjusts the fill light power according to the current external lighting status to achieve the purpose of stabilizing the lighting effect. This method optimizes and adjusts the target LED in the light source at a time, and cannot optimize the local lighting effect. The patent with application number 201520550995.9 controls the opening and closing of each LED through a switch, and adjusts the lighting effect of the target area by adjusting the relative position between the hardware. The adjustment process of this method is relatively complicated, and the optimization effect obtained is not necessarily the optimal solution. Therefore, the above research has certain limitations in the application of visual inspection systems. The paper "Research on Automatic Dimming Method of LED Shadowless Light Based on Image Tracking" (China Mechanical Engineering, 2012, 23(08):923-927) proposed an automatic real-time dimming method based on visual images. By analyzing the grayscale distribution of the image captured by the camera, the target area where the shadow needs to be eliminated is located. The shadow area is filled with light by controlling the fill light LED through the single-chip microcomputer to achieve real-time shadow elimination. This method can provide a higher average illumination for a larger area, but it needs to control 100 LEDs separately, and the control circuit is relatively complex, so there is still room for improvement. Therefore, designing a fill light system light source optimization method and device that can effectively reduce the requirements for LED hardware control circuits and LED brightness weight optimization algorithms while fully ensuring the lighting effect of the visual inspection system is of great significance to improving the lighting effect and inspection efficiency in machine vision inspection of industrial products. Summary of the invention

[0003] In order to overcome the above-mentioned defects, the present invention provides a method for optimizing the light source of a fill light system of a visual detection system and an experimental device used therein. The method for optimizing the light source of a fill light system of a visual detection system can enable the brightness weight of the fill light source to obtain the optimal lighting effect through a simple hardware control circuit and algorithm, thereby effectively improving the uniformity of illumination in the target area.

[0004] The present invention adopts a technical solution to solve the technical problem: a method for optimizing the light source of a fill light system of a visual inspection system, comprising the following steps:

[0005] Step 1: Install fill light source and external light source:

[0006] Install the fill light source bracket on the experimental table. The fill light source on the fill light source bracket provides fill light for the detection area. Install the external light source on one side of the experimental table to simulate the lighting environment under interference.

[0007] Step 2: Calculate the spatial pose parameters of each fill light source on the fill light source bracket;

[0008] Step 3: Establish a mathematical model of the fill light source. Through the mathematical model of a single fill light source and combined with the posture parameters of the current fill light source, obtain the illumination row vectors of the target area when each fill light source in the fill light system works alone, and arrange them in parallel into an illumination vector group.

[0009] Step 4: Analyze the illumination vector group of the fill light system through singular value decomposition, select the fill light sources corresponding to the first n illumination vectors with the largest principal element in the illumination vector group to replace all the fill light sources in the original model, and simplify the fill light system;

[0010] Step 5: Establish a plane illumination effect evaluation function based on illumination uniformity:

[0011] (1) Under the current lighting environment, the illumination distribution of the target area is sampled at equal intervals to obtain an n×n illumination matrix, and the rows of the illumination matrix are reordered from top to bottom into 1×n 2 The illumination row vector

[0012] (2) 1×n 2 The maximum value of the illumination vector of the target area under the action of the current external light source in the row vector of each fill light source is expressed as Subtract the illumination row vector Get the fill light illumination vector required by each fill light source

[0013]

[0014] (3) Vector It is expressed as the difference between the illumination distribution provided by the current fill light system and the currently required illumination distribution. The plane illumination effect evaluation function f is expressed as the trend similarity between the theoretical illumination distribution provided by the current fill light system and the currently required illumination distribution. i Represents the brightness weight of the i-th fill light source, ranging from [0, 1]. When the ratio of the current fill light source brightness to the maximum brightness is 1, it indicates the maximum brightness. is the illumination vector generated by the i-th LED on the target plane. The smaller f is, the closer the distribution patterns of the two are. At this time, the illumination distribution in the target area is more uniform:

[0015]

[0016]

[0017] Step 6: Apply the particle swarm algorithm to optimize and solve the plane illumination effect evaluation function to obtain the illumination weights of each LED in the fill light system with the optimal illumination effect;

[0018] Step 7: Adjust the brightness of the fill light source to optimize the lighting of the target area by the fill light system, so that the target area obtains a uniform lighting effect.

[0019] As a further improvement of the present invention, the fill light source includes LEDs arranged in a matrix on a curved surface and an Arduino control board. The Arduino control board controls the brightness of each LED in the fill light system by outputting corresponding PWM pulses. The external light source adopts a strip light source, and the strip light source is controlled by a light source controller to simulate a disturbed lighting environment.

[0020] As a further improvement of the present invention, in step 2: the method for calculating the spatial posture parameters of each fill light source is as follows:

[0021] Let (X0, Y0, Z0) be the base coordinate system, (X1, Y1, Z1) be the light source coordinate system, and set the position of an LED in the light source panel of the fill light system as the origin of the light source coordinate system;

[0022] After each coordinate system is established, the spatial pose parameters of the fill light source can be obtained from the relevant pose parameters of the bracket.

[0023] Assume dx, dy, and dz are the offsets of the origin of the light source coordinate system relative to the origin of the base coordinate system. The height difference between the current LED and the origin of the light source coordinate system in the fill light system is dz2, and the angle difference between the two relative to the origin in the XY plane of the base coordinate system is θ. Then the spatial position (x 01 ,y 01 ,z 01 ) is calculated as:

[0024]

[0025] As a further improvement of the present invention, the method for obtaining the illumination vector group in step 3 is as follows:

[0026] S31. Establish a mathematical model for a single LED light source:

[0027] A single LED can be regarded as a point light source with the same luminous flux, and its illumination distribution is determined by the cosine of the luminous angle:

[0028] E=I0·(cosα) m ·cosβ·d -2 (2)'

[0029] Where:

[0030] I0 is the average luminous intensity of the LED sphere;

[0031] m is the half-decay angle parameter of the LED. When the light source is an ideal Lambertian body, m is 1. Usually, m is greater than 1 and is determined by the half-decay angle θ of the current LED. 1 / 2 Sure;

[0032] α is the angle between the light ray and the optical axis;

[0033] β is the angle between the light and the normal of the measured plane;

[0034] d is the distance between the target point and the light source surface;

[0035] For a given LED, I0 and m are constants;

[0036] S32. Conversion and acquisition of variables:

[0037] Assume that the vector from each LED to the target point in the fill light system is a1, the normal vector of the plane where the light source is located is a2, the vertical vector in space is constant as G=(0,0,-1), the angle difference between the current LED and the origin of the light source coordinate system relative to the origin in the XY plane of the base coordinate system is θ, and L is the Euclidean distance from the light source point to the target point;

[0038] Then the calculation formulas for d, α, and β are:

[0039]

[0040] Among them, vector a1 is the coordinate of the measured point in the base coordinate system (x 02 ,y 02 ,z 02 ) and LED coordinates (x 01 ,y 01 ,z 01 ), the light source normal vector a2 can be obtained by θ i Calculate and obtain;

[0041]

[0042] The calculation formula of L is as follows:

[0043] L = |a1| (5);

[0044] S33. Obtain the illumination row vectors of the target area when each LED in the fill light system works alone through step S31 and step S32, and arrange them in parallel into an illumination vector group.

[0045] As a further improvement of the present invention, the method for simplifying the fill light system in step 4 is removed:

[0046] S41: performing singular value decomposition on the illumination vector group of the fill light system, calculating the size of the principal element of the illumination vector corresponding to each LED in the fill light system in the above vector group, comprehensively considering the diversity of the lighting effects of the fill light system and the reduction of the number of LEDs in the system, and determining the number of LEDs to be retained in the simplified fill light system;

[0047] S42: Determine the positions of the n required LEDs through Schmidt orthogonalization, use the corresponding n LEDs to replace the light source model in the original fill light system, and complete the simplification of the fill light system model. The specific steps are as follows:

[0048] set up is the illumination vector generated by the jth LED on the target plane, is the projection of the jth LED illumination vector on the orthogonal space axis vector during the nth feature vector screening process, The illumination vector corresponding to the LED retained in the simplified system is:

[0049] S421: Select illumination vector group The row vector with the largest modulus is taken as the first eigenvector and the illumination vector From the illumination vector group Removed;

[0050]

[0051] S422: Determine the second to nth eigenvectors according to Schmidt orthogonalization:

[0052] Calculate the projection vector set of all remaining illumination vectors on the orthogonal space axis vectors The calculation formula is as follows:

[0053]

[0054] S423: Maximum projection vector of modulus length The corresponding LED illumination vector Recorded as Then the illumination vector From the illumination vector group Removed, the calculation formula is as follows:

[0055]

[0056] S424: Repeat steps S422 and S423 until γ2 to γ n ;

[0057] S425: According to to Corresponding to the original illumination vector The position of n LEDs in the illumination matrix of the fill light system is obtained;

[0058] S43: Considering the impact of the simplification of the light source model of the fill light system on the average illumination, illumination uniformity, illumination distribution law, etc. of the system, the relative illumination calculation formula is designed as follows:

[0059]

[0060] Among them, e total_max The maximum value of the elements in the two illumination distribution matrices of the target area before and after the system optimization is obtained. The rationality of the simplification steps is proved by comparing the relative illumination distribution law of the target area under the simplified fill light system before and after.

[0061] As a further improvement of the present invention, in step 5, the lighting effect is optimized by minimizing the value of the evaluation function (12). k is the optimal solution for the brightness weight of each LED in the fill light system i The vector composed is:

[0062]

[0063] As a further improvement of the present invention, in step 5, the illumination distribution of the target area is sampled equidistantly under the current lighting environment to obtain a 9×9 illumination matrix, and the rows of the illumination matrix are reordered from top to bottom into 1×81 illumination row vectors.

[0064] As a further improvement of the present invention, in step seven, the PWM pulse corresponding to the optimized brightness weight is output by the Arduino control board to control the LED, so as to optimize the lighting of the target area by the fill light system, so that the target area obtains a uniform lighting effect. The specific steps are as follows:

[0065] S71: Obtain the conversion relationship between the analog value of the control board output port and the LED brightness weight:

[0066] Through testing, it is found that the relationship between the analog value of the control board output port and the current flowing through the LED is approximately linear. Combined with the fact that the LED brightness is proportional to the current, it is found that the analog value U of the control board output port and the LED brightness weight k are approximately linear.

[0067] In order to study the illumination distribution law of the target area when the fill light system is working, the brightness weight of each LED is obtained, and the LED brightness weight k is mapped to the analog quantity U at the output end of the control board. The specific calculation formula is as follows:

[0068] U = 255·k (14);

[0069] According to the theoretical and actual illumination distribution data of the target area, the relative illumination is calculated:

[0070]

[0071] Where, e(i,j) is the absolute illumination value of point (i,j), e single_max It is the maximum illumination value of the measured surface before and after optimization.

[0072] The difference between equation (15) and equation (9) is that the maximum value of the elements in the two illumination distribution matrices before and after optimization is used as the denominator. The denominator e in equation (15) is single_max is the maximum illumination value of the measured surface before and after optimization, e(i,j) is the absolute illumination value of point (i,j), and the effectiveness of the optimization target of the fill light system is proved by comparing the relative illumination distribution law of the target area before and after optimization.

[0073] An experimental device used in a light source optimization method for a fill light system of a visual inspection system comprises a test bench, a fill light source bracket, an LED, an Arduino control board, an illuminometer, an external light source, a light source controller and a processor. A target area for installing a product to be tested is formed on the test bench. The fill light source bracket is fixedly installed on the test bench at one side of the target area. A circular arc-shaped light source positioning plate extending in a vertical direction is fixedly installed on the fill light source bracket. A plurality of LEDs are evenly distributed on the concave surface of the light source positioning plate in a matrix shape. The LED light on the light source positioning plate can irradiate the product to be tested on the target area. The Arduino control board is electrically connected to each LED. The Arduino control board can output a PWM pulse signal to each LED. The external light source is fixedly installed on the test bench. The external light source can emit light simulating an actual inspection environment toward the target area. The light source controller controls the intensity of the light emitted by the external light source. The illuminometer is located on the test bench. The illuminometer can detect the illuminance value of a corresponding point on the target area. The illuminometer is electrically connected to the processor for communication. The processor can optimize and calculate the LED light source and control the Arduino control board to output a corresponding PWM pulse signal. The processor can also control the light source controller to emit a signal to the external light source.

[0074] As a further improvement of the present invention, the fill light source bracket includes a bracket body, a connecting block and a connecting block positioning piece. The bracket body extends vertically on the experimental bench in a direction perpendicular to the experimental bench. The connecting block can be slidably mounted on the bracket body in the longitudinal direction. The connecting block positioning piece can fix the connecting block of any height to the bracket body. The light source positioning plate is fixedly installed on the connecting block. Four rows and nine columns of LEDs are installed on the light source positioning plate. The origin of the base coordinate system is used as the center of the circle. The angle interval between adjacent columns of LEDs is 10°. The height difference between adjacent rows of LEDs is 30 mm. The normal vector of each LED is the normal of the tangent of the current position on the surface of the light source positioning plate. The external light source is a strip light source. The strip light source is installed on the experimental bench through a multi-axis arm. The multi-axis arm can change the position and inclination direction of the external light source in three-dimensional space.

[0075] The beneficial effects of the present invention are as follows: the present invention simulates the lighting environment under interference by installing an external light source on the experimental table, establishes a base coordinate system and a light source coordinate system with the same direction and different origin positions according to the fill light source bracket and the placement position under the condition of simulating the actual detection environment, realizes the mathematical model of LED, obtains the illumination vector group of LED in the fill light system, and establishes a plane illumination effect evaluation function after simplifying the fill light system by singular value decomposition, optimizes and solves the plane illumination effect evaluation function by using the particle swarm algorithm, and finally obtains the illumination weight of each LED in the fill light system when the optimal illumination effect is obtained. In the machine vision detection of industrial products, the present invention comprehensively considers the diversity of the lighting effects of the fill light system, greatly reduces the number of LEDs in the system, simplifies the fill light system model by singular value decomposition, and considers the fill light The illumination uniformity of the system light source illumination is obtained, and the brightness weights of each LED of the fill light system with the best illumination quality on the target plane are obtained, thereby improving the lighting effect of the target area. The present invention can obtain the optimal brightness weights of each LED of the fill light system, and the obtained theoretical illumination uniformity of the target area is about 92%, and the theoretical and actual optimized illumination distribution laws are basically the same. Compared with the existing fill light system light source optimization method, the present invention simplifies the fill light system model on the basis of fully ensuring the lighting effect of the visual inspection system, and only uses fewer LEDs to provide similar lighting effects, which effectively reduces the requirements for the LED hardware control circuit and the brightness weight optimization algorithm. The present invention can be used for the lighting link in the machine vision inspection of industrial products, provides better target product lighting effects for visual inspection, and improves the inspection efficiency of each link of subsequent visual inspection.

[0076] Instruction Manual

[0077] Figure 1 This is a flow chart of the light source optimization method for the fill light system proposed by the present invention;

[0078] Figure 2 This is a location diagram of the experimental device used for chip packaging quality inspection of the present invention;

[0079] Figure 3 This is a schematic diagram of the fill light system bracket and optimized area;

[0080] Figure 4 It is a schematic diagram of the fill light system bracket and various coordinate systems;

[0081] Figure 5 Schematic diagram of target vector and light source normal vector;

[0082] Figure 6 is the principal element value of each LED illumination vector;

[0083] Figure 7 To simplify the comparison diagram of the front and rear illumination distribution;

[0084] Figure 8This is a stereogram of the experimental device for the fill light system;

[0085] Fig. 9 It is the illumination grayscale relationship fitting diagram;

[0086] Fig.10 This is the relationship diagram between the port analog quantity and the LED current;

[0087] Fig.11 Theoretically optimized front and rear illumination distribution diagram when the external lighting is concentrated on P1;

[0088] Fig.12 To actually optimize the front and rear illumination distribution diagram when the exterior lighting is concentrated on P1;

[0089] Fig.13 Theoretically optimized front and rear illumination distribution diagram when the external lighting is concentrated at P2;

[0090] Fig.14 The front and rear illumination distribution diagram is actually optimized when the exterior lighting is concentrated on P2. DETAILED DESCRIPTION

[0091] The present invention is described in detail below in conjunction with the accompanying drawings. The embodiment described in the present invention is only a preferred embodiment of the present invention, rather than all embodiments. Based on the embodiment of the present invention, all other embodiments obtained by those skilled in the art without creative work are within the protection scope of the present invention.

[0092] Embodiment: A method for optimizing the light source of a supplementary light system of a visual inspection system. The flow chart of the optimization method is as follows: Figure 1 As shown, the specific steps include:

[0093] S1: Design the experimental device according to the optimization target. The optimization target of this embodiment is to detect defects on multiple chips in the tray loading mode. The target area is a 40mm×40mm square area, marked as follows Figure 2 As shown in the figure, the area covers a maximum of 4 rows and 4 columns, totaling 16 chips, so as to achieve the purpose of completing defect detection of up to 16 chips at a time. In the complete detection area covering all chips to be tested, the image acquisition and detection of chips in the non-overlapping area can be assisted by the relevant motion mechanism. Considering the diversity of the original lighting environment in the target area and the limitation of the spatial position of the devices around the visual inspection system on the size of the fill light system, the experimental device is designed.

[0094] The experimental device includes a test bench 1, a fill light source bracket, LED 2, an Arduino control board, an illuminance meter 3, an external light source 4, a light source controller and a processor. The test bench 1 is formed with a target area for installing the product to be tested. The bracket body 6 of the fill light source bracket is fixedly installed on the test bench 1. The connecting block 7 is adjusted to a suitable height, and the connecting block 7 is fixed with a connecting block 7 positioning piece. The light source positioning plate 5 is a 90° arc surface with the center of the target area as the center of the circle, and the radius of the arc surface is 100 mm. The light source positioning plate 5 is set at the upper right corner of the target area. 4 rows and 9 columns of LEDs 2 are installed on the concave arc surface of the light source positioning plate 5. The origin of the base coordinate system is the center of the circle, and the angle interval of adjacent columns of LEDs 2 is 10°. Figure 3 As shown; the height difference between adjacent rows of LED2 is 30mm, such as Figure 4 As shown, the normal vectors of each LED2 are the normals of the tangent line at the current position on the bracket surface.

[0095] A bar light source is used as the external light source 4, and the external light source 4 is installed on the experimental table 1 through a multi-axis arm 8. The external light source 4 is controlled by a light source controller to simulate a lighting environment under interference. The Arduino control board is used to output corresponding PWM pulses to control each LED 2 in the fill light system to improve the current lighting environment. The illuminance meter 3 is used to detect the illuminance value of the corresponding point in the target area. The illuminance meter 3 is electrically connected and communicated with the processor. The processor can optimize the calculation of the LED2 light source and control the Arduino control board to output the corresponding PWM pulse signal. The processor can also control the light source controller to transmit a signal to the external light source 4.

[0096] S2: According to the selected fill light source bracket and placement position, establish a base coordinate system and a light source coordinate system with the same direction and different origin positions to facilitate the calculation of the spatial position of each LED2 on the light source bracket:

[0097] Among them, (X0, Y0, Z0) is the base coordinate system, (X1, Y1, Z1) is the light source coordinate system, as shown in the following figure: Figure 4 In this embodiment, the position of LED2 in the 4th row and the 1st column of the light source positioning plate 5 of the supplementary light system is set as the origin of the light source coordinate system. At this time, the X axes of the two coordinate systems in the XY plane of the base coordinate system are collinear, and the three axes are in the same direction.

[0098] After each coordinate system is established, the spatial pose parameters of the fill light source can be obtained from the relevant pose parameters of the bracket. Among them, dx, dy, and dz are the offsets of the origin of the light source coordinate system relative to the origin of the base coordinate system. The height difference between the current LED and the origin of the light source coordinate system is dz2, and the angle difference between the two relative to the origin in the XY plane of the base coordinate system is θ. Then the spatial position (x 01 ,y 01 ,z 01 ) is calculated as shown in formula (1):

[0099]

[0100] S3: Based on the mathematical model of a single LED and the current LED posture parameters, the illumination row vectors of the target area when each LED in the fill light system works alone are obtained and arranged in parallel into an illumination vector group. The specific steps are as follows:

[0101] S31: Establishment of mathematical model of single LED light source:

[0102] A single LED can be regarded as a point light source with the same luminous flux, and its illumination distribution is determined by the cosine of the luminous angle:

[0103] E=I0·(cosα) m ·cosβ·d -2 (2);

[0104] Where I0 is the average luminous intensity of the LED sphere; m is a parameter related to the LED half-decay angle. When the light source is an ideal Lambertian body, m is 1. Usually m is greater than 1, and m is determined by the half-decay angle θ of the current LED. 1 / 2 Determine; α is the angle between the light and the optical axis; β is the angle between the light and the normal of the measured plane; d is the distance between the target point and the light source surface; corresponding to the given LED, I0 and m are constants.

[0105] S32: Conversion and acquisition of variables:

[0106] Assume that the vector from each LED in the fill light source to the target point is a1, the normal vector of the plane where the light source is located is a2, the vertical vector in space is constant as G=(0,0,-1), the angle difference between the current LED and the origin of the light source coordinate system relative to the origin in the XY plane of the base coordinate system is θ, and L is the Euclidean distance from the light source point to the target point. Figure 5 As shown, the calculation formulas for d, α, and β are:

[0107]

[0108] Among them, vector a1 can be obtained by the coordinates of the measured point in the base coordinate system (x 02 ,y 02 ,z 02 ) and LED coordinates (x 01 ,y 01 ,z 01 ), the light source normal vector a2 can be obtained by θ i Calculate and obtain;

[0109]

[0110] The calculation formula of L is as follows:

[0111] L = |a1| (5);

[0112] S33: Obtain illumination row vectors of the target area when each LED in the fill light system works alone through step S31 and step S32, and arrange them in parallel to form an illumination vector group.

[0113] S4: In order to adapt to different types of external lighting environments, the fill light system needs to control each LED individually and provide a variety of fill light solutions to achieve high-quality uniform lighting. In terms of the hardware of the fill light system, if a control board is used to independently control each LED in the fill light model, at least 36 pairs of analog output ports are required, which places high demands on the number of output ports of the control board; in terms of the optimization process, if the brightness weights of 36 LEDs are used as the optimization target, the dimension of the objective function is high, and it is difficult for the optimization algorithm to give a global optimal solution in a limited time. Therefore, it is considered to simplify the above fill light system model. The specific method is as follows:

[0114] S41: To prove that the simplification of the fill light system is theoretically feasible, the illuminance vector group of the fill light system is subjected to singular value decomposition, and the size of the principal element of the illuminance vector corresponding to each LED in the above vector group is calculated. The calculation results are as follows: Figure 6 As shown. Arrange the illumination vectors of each LED from large to small according to the principal element value, then the principal element value of the illumination vector corresponding to the 9th and subsequent LEDs in the above matrix is ​​basically 0, so each LED illumination vector can be approximately equivalent to a linear combination of the illumination vectors of the first 8 LEDs in the order. Therefore, considering the diversity of the lighting effects of the fill light system and reducing the number of LEDs in the system, the LEDs corresponding to the 8 illumination vectors with the largest principal element in the illumination matrix of the fill light system are selected to replace the 36 LEDs in the original model to simplify the fill light system.

[0115] S42: Determine the position parameters of the required 8 LEDs through Schmidt orthogonalization, use the corresponding 8 LEDs to replace the original fill light system light source model, and complete the simplification of the fill light system model. The specific steps are as follows:

[0116] First set is the illumination vector generated by the jth LED on the target plane, is the projection of the jth LED illumination vector on the orthogonal space axis vector during the nth feature vector screening process, The illumination vector corresponding to the LED retained in the simplified system.

[0117] S421: Select illumination vector group The row vector with the largest modulus is taken as the first eigenvector and the illumination vector From the illumination vector group Removed, the calculation formula is as follows:

[0118]

[0119] S422: Determine the 2nd to 8th eigenvectors according to Schmidt orthogonalization: Calculate the projection vector group of all remaining illumination vectors on the orthogonal space axis vectors The formula is shown in formula (7):

[0120]

[0121] S423: Maximum projection vector of modulus length The corresponding LED illumination vector Recorded as Then the illumination vector From the illumination vector group Remove from the equation (8):

[0122]

[0123] S424: Repeat steps S422 and S423 until γ2 to γ n

[0124] S425: According to to Corresponding to the original illumination vector The positions of the corresponding 8 LEDs are obtained from the positions in the illumination matrix of the fill light system.

[0125] Considering the impact of the simplification of the light source model of the fill light system on the system average illumination, illumination uniformity, illumination distribution law, etc., the relative illumination calculation formula is designed as shown in formula (9):

[0126]

[0127] where e total_max It is the maximum value of the elements in the two illumination distribution matrices of the target area before and after system optimization.

[0128] The illuminance distribution under the theoretical original model and the theoretical simplified model is as follows Figure 7As shown, XY corresponds to the XY coordinates of the target area in the base coordinate system, and the Z axis corresponds to the relative illumination. The illumination distribution diagram under the simplified model is compared with the illumination distribution diagram when the original 36 LEDs are working, which proves the effectiveness of the simplified model. The light-colored surface above is the theoretical relative illumination distribution diagram of the target area under the original scheme when 36 LEDs work together, and the dark-colored surface below is the theoretical illumination distribution diagram of the target area under the simplified scheme when 8 LEDs are working after simplification. It can be seen that the illumination distribution rules of the two are relatively similar. Since the number of LEDs is reduced to 8 after simplification, this process discards the LEDs corresponding to the illumination vectors with smaller principal elements in the illumination matrix. Therefore, compared with the original model, the maximum value of the illumination distribution that the simplified model can provide is reduced. However, on the basis of fully ensuring the lighting effect of the visual inspection system, the simplified model only uses fewer LEDs to provide similar lighting effects, which can effectively reduce the requirements of the LED hardware control circuit and the brightness weight optimization algorithm.

[0129] In order to verify that the number of LEDs retained in the simplified steps of the fill light system model should not be less than 8, the number of LEDs retained in the system is adjusted downward to perform statistical calculations of the mean square error of illumination. Figure 2 As shown in the figure, the biased external light source (strip LED) is regarded as the original lighting environment of the visual system under interference. Taking the case where the external light source is concentrated at points P1 = (-20, -20) and P2 = (-20, 0) in the XY plane of the base coordinate system as an example, the target area is sampled equidistantly to obtain a 9×9 illumination matrix, and the mean square error of the illumination distribution before and after theoretical optimization is calculated in the corresponding case. The mean square error of the illumination distribution before and after theoretical optimization when the number of retained LEDs is 5 to 8 is shown in Table 1.

[0130] Table 1 Mean square error of illumination distribution before and after theoretical optimization for different numbers of LEDs

[0131]

[0132] From the analysis of Table 1, it can be seen that when the number of LEDs in the supplementary lighting system is less than 8, the mean square error of the illumination distribution obtained by theoretical optimization of the target area is large. When 8 LEDs are selected, the mean square error of the illumination distribution obtained by theoretical optimization of the target area is the smallest, and the illumination distribution tends to be more uniform. Therefore, it can be verified that the solution of simplifying the original system model when 36 LEDs are in action to the simplified system model when 8 specific LEDs are in action is reasonable and effective.

[0133] S5: Establish a plane illumination effect evaluation function based on illumination uniformity, specifically including:

[0134] S51: Sample the illumination distribution of the target area at equal intervals under the current lighting environment to obtain a 9×9 illumination matrix, and reorder the rows of the illumination matrix from top to bottom into a 1×81 illumination row vector

[0135] S52: Set Each element in the 1×81 row vector is the illumination vector of the target area under the current external light source. The maximum value. Subtract the illumination row vector Get the required fill light illumination vector

[0136]

[0137] S53: Let vector It represents the difference between the illumination distribution provided by the current fill light system and the currently required illumination distribution, as shown in formula (11):

[0138]

[0139] The plane illumination effect evaluation function f is used to represent the trend similarity between the theoretical illumination distribution provided by the current fill light system and the current actual required illumination distribution, as shown in formula (12):

[0140]

[0141] Among them, k i represents the brightness weight of the i-th LED, ranging from [0, 1]. The brightness weight represents the ratio of the current LED brightness to the maximum brightness. When f is 1, it represents the maximum brightness. The smaller f is, the closer the distribution of the two is, and the more uniform the illumination distribution in the target area is:

[0142] S54: To achieve the best lighting effect, the value of the evaluation function (12) should be minimized. k is the optimal solution for the brightness weight of each LED in the fill light system i The vector composed is:

[0143]

[0144] S6: Apply the particle swarm algorithm to optimize and solve the plane illumination effect evaluation function to obtain the illumination weight of each LED of the fill light system with the optimal illumination effect.

[0145] S7: The PWM pulse corresponding to the optimized brightness weight output by the Arduino control board is used to control the LED to optimize the lighting of the target area by the fill light system, so that the target area obtains a uniform lighting effect.

[0146] Furthermore, in order to obtain the LED output parameters and verify the effectiveness of the supplementary lighting solution in this paper, the following Figure 8The experimental bench is shown in the figure. The external light source uses JH-LS6222 bar light source and JH-AP-2C light source controller from Jutu Optoelectronics. The fill light system is 8 LEDs of the same size, and the brightness of each LED is controlled by outputting 8 PWM pulses through the Arduino control board. The camera uses Invision OSR500-20GM. The light source illumination is measured using Taiwan TES-1332A illuminometer, with a measurement range of 0-20000Lux and a measurement accuracy of 1Lux.

[0147] Further, according to Figure 8 The experimental environment shown in the figure is used to determine the quantitative conversion relationship between image grayscale and actual illumination. The external light source power is adjusted by the light source controller, and 10 sets of illumination and corresponding grayscale data are recorded. The fitting results are shown in Fig. 9 As shown, it can be seen that the two are approximately linearly related in a certain range.

[0148] Furthermore, in step S7, the specific steps for determining the magnitude of each LED driving pulse analog quantity are as follows:

[0149] (1) The 9×9 grayscale matrix of the corresponding points of the image under the current lighting environment is converted into an illumination matrix according to the grayscale-illuminance conversion relationship, and the rows of the illumination matrix are reordered from top to bottom into 1×81 illumination row vectors.

[0150] (2) According to the illumination vector under the current lighting environment The particle swarm algorithm is used to optimize the brightness weights of each LED in the fill light system, and the brightness weights of each LED in the fill light system are obtained when the illumination distribution in the target area is the most uniform.

[0151] (3) Adjust the analog value output from the output port of the control board and record the LED current at this time. After testing, the analog value of the output port of the control board is basically linearly related to the current flowing through the LED, such as Fig.10 As shown. Because the LED brightness is proportional to the current, the analog quantity U at the output port of the control board and the LED brightness weight k are approximately linearly related.

[0152] In order to study the illumination distribution law of the target area when the fill light system is working, after obtaining the brightness weight of each LED in step (2), the LED brightness weight k can be mapped to the analog quantity U at the output end of the control board, as shown in formula (14):

[0153] U = 255·k (14);

[0154] Since the fill light system is installed in Figure 3In the upper right corner of the target area shown in the figure, in order to highlight the optimization effect and reflect the working performance of the fill light system under different lighting environments, experimental verification and analysis are carried out for the two cases where the external light source illumination center is concentrated in P1 and P2. The particle swarm algorithm is used to optimize the evaluation function shown in formula (12) to obtain the LED brightness weight vector corresponding to the optimal value of the illumination effect evaluation function of the measured area, where is the optimal brightness weight vector of the LED of the fill light system when the external lighting is concentrated at point P1, It is the optimal brightness weight vector of the LED of the fill light system when the external lighting is concentrated at point P2.

[0155]

[0156]

[0157] The eight values ​​in the vector correspond to the brightness weights of the eight LEDs at different positions in the fill light system. Since the original illumination distribution of the target area in the experiment is uneven, the brightness of the LEDs at different positions needs to change dynamically to improve the overall illumination uniformity of the target area. If the evaluation function value decreases due to the current LED being turned on, that is, the beneficial effect of improving the lighting environment is generated, then the brightness weight of the LED is positive; otherwise, the brightness weight of the LED is 0. According to the optimal brightness weight of each LED under the current situation, the analog value of the corresponding output end of the control board is calculated and set in combination with formula (14), and the brightness of each LED is adjusted by outputting PWM through the control board. The image of the optimized target area is captured to obtain the actual illumination distribution of the optimized target area.

[0158] The relative illumination is calculated based on the theoretical and actual illumination distribution data of the target area, as shown in formula (15):

[0159]

[0160] The difference between equation (15) and equation (9) is that the maximum value of the elements in the two illumination distribution matrices before and after optimization is used as the denominator. The denominator e in equation (15) is si ng l e_max is the maximum illumination value of the measured surface before and after optimization, and e(i,j) is the absolute illumination value of point (i,j).

[0161] The comparison of theoretical and actual illuminance distribution before and after optimization for the two cases where external lighting is concentrated on P1 and P2 is shown in the figure below. Fig.11 , Fig.12 , Fig.13 and Fig.14 As shown, XY corresponds to the XY coordinates of the target area in the base coordinate system, and the Z axis corresponds to the relative illumination value.

[0162] Fig.11 and Fig.13 The original illuminance distribution of the target area and the illuminance distribution after theoretical optimization are shown when the illumination center of the external light source is concentrated at P1 and P2. The light color is the original illuminance distribution, and the dark color is the illuminance distribution after theoretical optimization. It can be seen that the fill light system can theoretically improve the lighting effect of the target area.

[0163] Fig.12 and Fig.14 The original illuminance distribution of the target area when the external lighting is concentrated on P1 and P2 and the actual illuminance distribution after optimization. The light color is the original illuminance distribution, and the dark color is the actual illuminance distribution after optimization. It can be seen that the fill light system can also improve the uniformity of the illumination distribution in the target area in the actual optimization, and the theoretical and actual illuminance distribution rules of the target area after optimization are basically the same.

[0164] Calculate separately Fig.11 , Fig.12 , Fig.13 and Fig.14 Illumination uniformity f uni , as shown in (16):

[0165]

[0166] where e min is the minimum value of the current illumination vector, is the current illumination vector mean. The original illumination uniformity, theoretically optimized illumination uniformity, and actual optimized illumination uniformity corresponding to the external lighting concentrated at P1 and P2 are shown in Table 2.

[0167] Table 2 Illuminance uniformity correspondence table

[0168]

[0169] As can be seen from Table 2, in the two groups of experiments where the external lighting is concentrated on P1 and P2, the light source optimization method for the fill light system proposed in this application can theoretically increase the illumination uniformity of the target area to about 92%, and in the actual system, it can increase the illumination uniformity of the target area to 84.54% and 74.05%, respectively, which are 35.51% and 14.09% higher than before optimization, respectively, proving the effectiveness of the light source optimization system of this application. Since this application studies the illumination distribution law of the target area when the fill light system is working, and does not quantitatively control the illumination distribution amplitude, there are differences in the illumination distribution of the target area after theoretical and actual optimization, but the illumination distribution law of theoretical and actual optimization is basically the same, proving the feasibility of the light source optimization method for the fill light system. Comparing the actual optimization results under the above two different positions, when the external lighting is concentrated on P1, since the average illumination of the lighting environment provided by the fill light system is closer to the actual required uniform lighting effect, the actual fill light effect that the system can achieve under the corresponding situation is better.

[0170] In summary, the image-based fill light system proposed in this application can still effectively improve the lighting effect of the target area based on a simplified model, and the theoretical optimized illumination distribution is basically consistent with the actual optimized illumination distribution law. Therefore, the fill light system light source optimization method proposed in this application is reasonable and effective.

[0171] The above is only a preferred specific embodiment of the present invention, but the protection scope of the present invention is not limited thereto. Any changes or substitutions that can be easily thought of by a person skilled in the art within the technical scope disclosed in the present invention should be included in the protection scope of the present invention. Therefore, the protection scope of the present invention should be based on the protection scope of the claims.

Claims

1. A method for optimizing the light source of a fill light system of a visual inspection system, characterized in that: The following steps are involved: Step 1: Install fill light source and external light source: Install the fill light source bracket on the experimental table. The fill light source on the fill light source bracket provides fill light for the detection area. Install the external light source on one side of the experimental table to simulate the lighting environment under interference. Step 2: Calculate the spatial pose parameters of each fill light source on the fill light source bracket; Step 3: Establish a mathematical model of the fill light source. Through the mathematical model of a single fill light source and combined with the posture parameters of the current fill light source, obtain the illumination row vectors of the target area when each fill light source in the fill light system works alone, and arrange them in parallel into an illumination vector group. Step 4: Analyze the illumination vector group of the fill light system through singular value decomposition, select the fill light sources corresponding to the first n illumination vectors with the largest principal element in the illumination vector group to replace all the fill light sources in the original model, and simplify the fill light system; Step 5: Establish a plane illumination effect evaluation function based on illumination uniformity: (1) Under the current lighting environment, the illumination distribution of the target area is sampled at equal intervals to obtain an n×n illumination matrix, and the rows of the illumination matrix are reordered from top to bottom into 1×n 2 The illumination row vector (2) 1×n 2 The maximum value of the illumination vector of the target area under the action of the current external light source in the row vector of each fill light source is expressed as Subtract the illumination row vector Get the fill light illumination vector required by each fill light source (3) Vector It is expressed as the difference between the illumination distribution provided by the current fill light system and the currently required illumination distribution. The plane illumination effect evaluation function f is expressed as the trend similarity between the theoretical illumination distribution provided by the current fill light system and the currently required illumination distribution. i Represents the brightness weight of the i-th fill light source, ranging from [0, 1]. When the ratio of the current fill light source brightness to the maximum brightness is 1, it indicates the maximum brightness. is the illumination vector generated by the i-th LED on the target plane. The smaller f is, the closer the distribution patterns of the two are. At this time, the illumination distribution in the target area is more uniform: Step 6: Apply the particle swarm algorithm to optimize and solve the plane illumination effect evaluation function to obtain the illumination weights of each LED in the fill light system with the optimal illumination effect; Step 7: Adjust the brightness of the fill light source to optimize the lighting of the target area by the fill light system, so that the target area obtains a uniform lighting effect.

2. The method for optimizing the light source of the fill light system of the visual inspection system according to claim 1, characterized in that: The fill light source includes LEDs arranged in a matrix on a curved surface and an Arduino control board. The Arduino control board controls the brightness of each LED in the fill light system by outputting corresponding PWM pulses. The external light source adopts a strip light source, and the strip light source is controlled by a light source controller to simulate a disturbed lighting environment.

3. The method for optimizing the light source of the supplementary light system of the visual inspection system according to claim 2, characterized in that: In step 2: the spatial pose parameters of each fill light source are calculated as follows: Let (X0, Y0, Z0) be the base coordinate system, (X1, Y1, Z1) be the light source coordinate system, and set the position of an LED in the light source panel of the fill light system as the origin of the light source coordinate system; Assume dx, dy, and dz are the offsets of the origin of the light source coordinate system relative to the origin of the base coordinate system. The height difference between the current LED and the origin of the light source coordinate system in the fill light system is dz2, and the angle difference between the two relative to the origin in the XY plane of the base coordinate system is θ. Then the spatial position (x 01 ,y 01 ,z 01 ) is calculated as:

4. The method for optimizing the light source of the supplementary light system of the visual inspection system according to claim 3, characterized in that: The method for obtaining the illumination vector group in step 3 is as follows: S31. Establish a mathematical model for a single LED light source: E=I0·(cosα) m ·cosβ·d -2 (2) Where: I0 is the average luminous intensity of the LED sphere; m is the half-decay angle parameter of the LED. When the light source is an ideal Lambertian body, m is 1. Usually, m is greater than 1 and is determined by the half-decay angle θ of the current LED. 1 / 2 Sure; α is the angle between the light ray and the optical axis; β is the angle between the light and the normal of the measured plane; d is the distance between the target point and the light source surface; For a given LED, I0 and m are constants; S32. Conversion and acquisition of variables: Assume that the vector from each LED to the target point in the fill light system is a1, the normal vector of the plane where the light source is located is a2, the vertical vector in space is constant as G=(0,0,-1), the angle difference between the current LED and the origin of the light source coordinate system relative to the origin in the XY plane of the base coordinate system is θ, and L is the Euclidean distance from the light source point to the target point; Then the calculation formulas for d, α, and β are: Among them, vector a1 is the coordinate of the measured point in the base coordinate system (x 02 ,y 02 ,z 02 ) and LED coordinates (x 01 ,y 01 ,z 01 ), the light source normal vector a2 can be obtained by θ i Calculate and obtain; The calculation formula of L is as follows: L = |a1| (5); S33. Obtain the illumination row vectors of the target area when each LED in the fill light system works alone through step S31 and step S32, and arrange them in parallel into an illumination vector group.

5. The method for optimizing the light source of the supplementary light system of the visual inspection system according to claim 2 or 4, characterized in that: The method for simplifying the fill light system in step 4 is removed: S41: performing singular value decomposition on the illumination vector group of the fill light system, calculating the size of the principal element of the illumination vector corresponding to each LED in the fill light system in the above vector group, comprehensively considering the diversity of the lighting effects of the fill light system and the reduction of the number of LEDs in the system, and determining the number of LEDs to be retained in the simplified fill light system; S42: Determine the positions of the n required LEDs through Schmidt orthogonalization, use the corresponding n LEDs to replace the light source model in the original fill light system, and complete the simplification of the fill light system model. The specific steps are as follows: set up is the illumination vector generated by the jth LED on the target plane, is the projection of the jth LED illumination vector on the orthogonal space axis vector during the nth feature vector screening process, The illumination vector corresponding to the LED retained in the simplified system is: S421: Select illumination vector group The row vector with the largest modulus is taken as the first eigenvector and the illumination vector From the illumination vector group Removed; S422: Determine the second to nth eigenvectors according to Schmidt orthogonalization: Calculate the projection vector set of all remaining illumination vectors on the orthogonal space axis vectors The calculation formula is as follows: S423: Maximum projection vector of modulus length The corresponding LED illumination vector Recorded as Then the illumination vector From the illumination vector group Removed, the calculation formula is as follows: S424: Repeat steps S422 and S423 until the S425: According to to Corresponding to the original illumination vector The position of n LEDs in the illumination matrix of the fill light system is obtained; S43: Considering the impact of the simplification of the light source model of the fill light system on the average illumination, illumination uniformity and illumination distribution of the system, the relative illumination calculation formula is designed as follows: Among them, e total_max It is the maximum value of the elements in the two illumination distribution matrices of the target area before and after system optimization.

6. The method for optimizing the light source of the supplementary light system of the visual inspection system according to claim 1, characterized in that: In step 5, the lighting effect is optimized by minimizing the value of the evaluation function (12). k is the optimal solution for the brightness weight of each LED in the fill light system i The vector composed is:

7. The method for optimizing the light source of the supplementary light system of the visual inspection system according to claim 6, characterized in that: In step 5, the illumination distribution of the target area is sampled equidistantly under the current lighting environment to obtain a 9×9 illumination matrix, and each row of the illumination matrix is ​​reordered from top to bottom into a 1×81 illumination row vector 8. The method for optimizing the light source of the supplementary light system of the visual inspection system according to claim 6, characterized in that: In step 7, the PWM pulse corresponding to the optimized brightness weight is output by the Arduino control board to control the LED, so as to optimize the lighting of the target area by the fill light system, so that the target area can obtain a uniform lighting effect. The specific steps are as follows: S71: Obtain the conversion relationship between the analog value of the control board output port and the LED brightness weight: Through testing, it is found that the relationship between the analog value of the control board output port and the current flowing through the LED is approximately linear. Combined with the fact that the LED brightness is proportional to the current, it is found that the analog value U of the control board output port and the LED brightness weight k are approximately linear. The LED brightness weight k is mapped to the analog quantity U at the output of the control board. The specific calculation formula is as follows: U = 255·k(14); According to the theoretical and actual illumination distribution data of the target area, relative illumination calculation is performed: Where, e(i,j) is the absolute illumination value of point (i,j), e single_max It is the maximum illumination value of the measured surface before and after optimization.

9. An experimental device used in the light source optimization method of the supplementary light system of the visual inspection system as claimed in claim 1, characterized in that: The invention comprises a test bench (1), a fill light source bracket, an LED (2), an Arduino control board, an illuminometer (3), an external light source (4), a light source controller and a processor. The test bench is formed with a target area for installing a product to be tested. The fill light source bracket is fixedly mounted on the test bench at one side of the target area. A circular arc-shaped light source positioning plate (5) extending in a vertical direction is fixedly mounted on the fill light source bracket. A plurality of LEDs are evenly distributed on the concave surface of the light source positioning plate in a matrix shape. The LED light on the light source positioning plate can irradiate the product to be tested on the target area. The Arduino control board is electrically connected to each LED. The Arduino control board can output a PWM pulse signal to each LED. The external light source is fixedly mounted on the test bench. The external light source can emit light simulating an actual detection environment toward the target area. The light source controller controls the intensity of the light emitted by the external light source. The illuminometer is located on the test bench. The illuminometer can detect the illuminance value of a corresponding point on the target area. The illuminometer is electrically connected to the processor for communication. The processor can optimize the calculation of the LED light source and control the Arduino control board to output a corresponding PWM pulse signal. The processor can also control the light source controller to emit a signal to the external light source.

10. The experimental device according to claim 9, characterized in that: The fill light source bracket comprises a bracket body (6), a connecting block (7) and a connecting block positioning piece. The bracket body extends vertically on the experimental table along a direction perpendicular to the experimental table. The connecting block can be slidably mounted on the bracket body along the longitudinal direction. The connecting block positioning piece can fix and position the connecting block of any height with the bracket body. The light source positioning plate is fixedly mounted on the connecting block. Four rows and nine columns of LEDs are mounted on the light source positioning plate. The origin of the base coordinate system is taken as the center of the circle. The angle interval of adjacent columns of LEDs is 10°. The height difference of adjacent rows of LEDs is 30 mm. The normal vectors of each LED are the normals of the tangents at the current position on the curved surface of the light source positioning plate. The external light source is a strip light source. The strip light source is mounted on the experimental table via a multi-axis arm (8). The multi-axis arm can change the position and tilt direction of the external light source in three-dimensional space.

Images