Polarized light-based defect detection apparatus and method for flexible OLED panel

Abstract

A polarized light-based defect detection apparatus for a flexible OLED panel, comprising: a sample conveying apparatus, comprising a conveyor belt (7); an acquisition apparatus, comprising a detector (1) and a light source (4); a polarization apparatus comprising an analyzer (2), a quarter-wave plate II (3), a polarizer (5), and a quarter-wave plate I (6), and being used for acquiring polarization information of a flexible OLED panel sample to be tested (11); an angle changing apparatus, comprising an electric rotary table (9) and a dial (8); and a computer (10) used for acquiring light intensity values and power values of each point of said flexible OLED panel sample (11) by means of the acquisition apparatus at different moments, and changing angles at the different moments to obtain multiple groups of light intensity values and power values, so as to calculate different parameters and acquire different parameters of a detect type and the position of a layer having a defect. Also disclosed is a polarized light-based defect detection method for a flexible OLED panel.

Description

A flexible OLED panel defect detection device and method based on polarized light TECHNICAL FIELD

[0001] The present application relates to the field of workpiece manufacturing and processing, in particular to a flexible OLED panel defect detection device and method based on polarized light. BACKGROUND

[0002] A flexible OLED is composed of a flexible substrate, a flexible electrode, a transmission layer and a light-emitting material. Complex production processes are prone to cause various display defects, and various defects are prone to react with organic materials to cause rapid aging of the device, which has a great impact on the performance of the flexible screen. Different defects of different layers need to be detected differently, and different solutions can be provided according to the type and location of the defects. Therefore, high-quality detection of defects can effectively eliminate unqualified products, provide repair basis for repairable defects, trace the causes of defects, and further improve the production process and increase the yield.

[0003] Traditional manual detection has low efficiency and cannot meet the needs of assembly line production. In addition, it is also affected by factors such as fatigue and subjective consciousness, resulting in false detection, missed detection and other shortcomings. Machine vision detection is to use optical equipment such as machines to replace the human eye for measurement and judgment. Usually, image acquisition equipment is used to convert the research object into corresponding digital signals, and the converted signal characteristics are extracted and distinguished, and the next device operation is based on the analysis results. In the past few decades, machine vision technology has gradually matured and is widely used in manufacturing. Compared with manual detection methods, defect detection methods relying on machine vision can achieve rapid defect detection in a flow line, reduce costs, liberate labor, and improve detection accuracy and efficiency. At present, many domestic and foreign experts and scholars have made a lot of research on defect detection based on machine vision and have achieved good results, but there is still a lot of research space in the accuracy and speed of flexible OLED panel defect detection. Research on flexible OLED display screen defect detection not only has certain application prospect, but also can further promote the development of flexible OLED display panel industry.

[0004] When defect detection is performed on a flexible OLED panel with high surface transparency, distortion caused by refraction and scattering of light often accompanies, polarized light can weaken the influence of glare or bright spots from the reflecting surface, enhance image contrast, and eliminate strong reflection. In addition, defects generated by different layers of the flexible OLED panel need to be detected, and different processing is required for defects appearing in different layers. Feature extraction using a gray intensity map often cannot distinguish the position of the defect between layers, while the bidirectional reflectance distribution function of the p-polarized scattered light caused by the p-polarized incident light can determine the layer position of the defect. Determining the type, position and layer position of the defect can better provide a basis for production process improvement.

[0005] Summary of the application

[0006] The application builds an optical system to collect images of flexible OLED panel samples on the assembly line, designs algorithms based on various parameters of polarized light, and trains a classifier algorithm to achieve defect detection.

[0007] The application is achieved by the following technical solutions:

[0008] The application discloses a flexible OLED panel defect detection device based on polarized light, comprising:

[0009] The sample conveying device comprises a conveying belt, which is used to convey the flexible OLED panel sample to be tested to the collection area and convey the tested flexible OLED panel sample to the subsequent station after detection;

[0010] The collection device comprises a detector and a light source, the light source is used to light the surface of the tested flexible OLED panel sample, and the detector is used to obtain the light intensity value and power value of each point of the tested flexible OLED panel sample;

[0011] The polarization device comprises an analyzer, a second 1 / 4 wave plate, a polarizer and a first 1 / 4 wave plate, and is used to obtain the polarization information of the tested flexible OLED panel sample;

[0012] The angle conversion device comprises a motorized rotary table and a scale disc, the motorized rotary table is used to control the angle conversion of the polarization device, and the scale disc is used to control the angle conversion of the collection device;

[0013] The computer is used to synchronously control the collection speed of the collection device and the movement speed of the sample conveying device, to collect the light intensity value and power value of each point of the tested flexible OLED panel sample at different times; to control the angle conversion of the angle conversion device to change the angle to obtain multiple sets of light intensity value and power value, to realize the calculation of different parameters; to control the switching of the use state of the first 1 / 4 wave plate and the second 1 / 4 wave plate and the switching of the type of the light source, to realize the acquisition of different parameters of the defect type and the defect level position;

[0014] The dial includes a dial shaft support A and a dial shaft support B, the dial shaft support A is fixed with a detector, an analyzer and a second 1 / 4 wave plate from top to bottom, the dial shaft support B is fixed with a light source, a polarizer and a first 1 / 4 wave plate from top to bottom, and the computer is connected with the motorized rotary table, the conveying belt, the dial and the light source respectively.

[0015] The collection area is that the flexible OLED panel sample to be measured is located in the field of view of the detector and the illumination range of the light source at the same time.

[0016] As a further improvement, the rotation of the angle of the analyzer, the second 1 / 4 wave plate, the polarizer, the first 1 / 4 wave plate in the device is realized by the rotation of the motorized rotary table below, the rotation of the angle of the incident light is realized by the rotation of the dial shaft support B, the rotation of the angle of the scattered light is realized by the rotation of the dial shaft support A, and the rotation of the motorized rotary table, the dial shaft support A and the dial shaft support B is realized by the computer control.

[0017] The application also includes a flexible OLED panel defect detection method based on polarized light, including the following steps:

[0018] 1) Switch the light source to a white light source, control the conveying belt to move until the standard sample is placed in the measurement range, and obtain the Stokes vector of the defect area and the defect-free area under a plurality of experimental angles;

[0019] 2) Calculate the Mueller matrix parameters of the defect area and the defect-free area from the Stokes vector, select the optimal angle state from a plurality of angles, and train the classifier using the Mueller matrix parameter value under the optimal angle state to obtain a trained classifier;

[0020] 3) Adjust the angle of the dial shaft support B, the polarizer and the first 1 / 4 wave plate to the optimal angle state, start the conveying belt, control the conveying belt to move until the flexible OLED panel sample to be measured is placed in the measurement range, control the motorized rotary table, and obtain the Stokes vector of each pixel point of the flexible OLED panel sample to be measured;

[0021] 4) Calculate the Mueller matrix parameters of each pixel point of the flexible OLED panel sample to be measured from the Stokes vector of each pixel point, input the trained classifier for identification and detection, and obtain the defect type and the pixel point position;

[0022] 5) Control the light source to switch to a laser, the first 1 / 4 wave plate and the second 1 / 4 wave plate to switch to a non-use state, adjust the angle of the dial shaft support B to a fixed value, adjust the receiving angle of the scattered light in the angle range by the dial shaft support A, and obtain the power value measured by the detector;

[0023] 6) Calculate the change trend of the polarization factor of each pixel point of the flexible OLED panel sample to be tested by the power value, analyze the defect position belonging to the surface, the subsurface or the rough surface, and obtain the hierarchical position of the flexible OLED panel where the defect is located;

[0024] 7) The hierarchical position of the flexible OLED panel is fused as supplementary information into the defect type and the pixel point position result, and the comprehensive result of the defect detection is obtained as the conclusion output.

[0025] As a further improvement, the plurality of experimental angles in step 1) of the present application are that the incident light angle is in the range of 0° to 80°, and every 10° is a group of selected incident light angles; the PSG angle is in the range of 0° to 180°, and every 10° is a group of selected PSG angles.

[0026] The fixed value of the incident light angle in step 5) is 60°, and the scattering light receiving angle range is 36°-72°.

[0027] As a further improvement, the steps for calculating the Stokes vector in steps 1) and 3) of the present application are:

[0028] 1) Record the light intensity values under six combinations of (0°, 0°), (0°, 45°), (45°, 45°), (90°, 90°), (0°, 135°), and (135°, 135°), and record the light intensity values of each pixel point obtained under the six angle combination conditions as I1, I2, I3, I4, I5, and I6, respectively;

[0029] 2) Calculate the Stokes vector S = [A1, A2, A3, A4] of each pixel point by combining the light intensity values of each pixel point with the formulas A1 = (I1 + I4), A2 = (I1 - I4), A3 = (I3 - I6), and A4 = (I5 - I2).

[0030] As a further improvement, the calculation of the Mueller matrix parameter in step 2) of the present application is as follows: let the incident light Stokes vector be S and the outgoing light Stokes vector be S', and the Mueller matrix parameter M of each pixel point can be calculated according to the formula S' = M·S. The M parameter is a four-row and four-column matrix, and the value of the first row and the first column is taken as the parameter value as the input data of the training classifier.

[0031] The calculation of the Mueller matrix parameter in step 4) is as follows: let the incident light Stokes vector be S and the outgoing light Stokes vector be S', and the Mueller matrix parameter M of each pixel point can be calculated according to the formula S' = M·S. The M parameter is a four-row and four-column matrix, and the deviation of the value of the first row and the first column from the parameter value of the defect-free area under the same angle condition of the standard sample is taken as the input data of the detection classifier.

[0032] As a further improvement, the step 2) of the present application is the selection step of the optimal angle state:

[0033] 1) Calculate the Mueller matrix parameters of each type of defect region and the Mueller matrix parameters of the defect-free region under each incident light angle and each PSG angle;

[0034] 2) Calculate the deviation percentage of each type of defect region compared to the defect-free region based on the Mueller matrix parameters of the defect-free region;

[0035] 3) Set the deviation percentage equal to 30% as the reference line for distinguishing between defect-free regions and defect regions, and 50% as the reference line for distinguishing between different defect regions, select the incident light angle and PSG angle that meet the distinction criteria, if there are multiple sets of angles that meet the criteria, select the set with the largest deviation as the optimal angle state, if there is only one set of angles that meet the criteria, select that set as the optimal angle state.

[0036] As a further improvement, the step 2) of the present application is the training step of the classifier:

[0037] 1) Adjust the SVM support vector machine classifier to a classifier that needs to distinguish between the number of defect types to be detected plus one (i.e. if there are N types of defects, adjust the SVM support vector machine to a (N+1) classifier);

[0038] 2) Perform multiple sets of repetitive experiments on the optimal angle state in the adjusted classifier, calculate the multiple sets of Mueller matrix parameter deviation values as input data sets for the training of the classifier, and save the trained classifier model for future use.

[0039] As a further improvement, the step 6) of the present application is the change trend calculation step of the polarization factor:

[0040] 1) Convert the power value to BRDF pp ;

[0041] 2) Draw the BRDF pp change curve to obtain the trend of BRDF pp change with the scattering light angle.

[0042] As a further improvement, the step 6) of the present application is the judgment basis for whether the defect position belongs to the surface above, subsurface or micro-rough surface:

[0043] 1) If the BRDF pp increases with the increase of the scattering azimuth angle, it means that the defect is above the surface;

[0044] 2) If the BRDF ppThe minimum value appears near the scattering azimuth angle of 36°, and the BRDF in other ranges pp The change with the scattering azimuth angle indicates that the defect is in the micro-rough surface;

[0045] 3) If the BRDF pp The minimum value appears near the scattering azimuth angle of 72°, and the BRDF in other ranges pp The change with the scattering azimuth angle indicates that the defect is in the subsurface.

[0046] The beneficial effects of the present application are as follows:

[0047] 1. The present application can quickly scan the flexible OLED panel sample to be tested, comprehensively detects the flexible OLED panel sample to be tested, and significantly improves the detection efficiency;

[0048] 2. The present application combines advanced image processing algorithms, can automatically identify and classify each defect of the flexible OLED panel sample to be tested, and reduces the labor cost and time cost.

[0049] 3. The present application can identify various defects on the surface of the flexible OLED panel by calculating and comparing the differences of Mueller matrixes of different defects, and widen the defect detection range.

[0050] 4. The present application can accurately locate the position of the layer where the defect of the flexible OLED panel sample to be tested is located by calculating the change trend of the polarization factor of each pixel point.

[0051] 5. The present application uses a visual detection scheme, calculates the defect information by a computer, and avoids damage to the flexible OLED panel sample to be tested in the detection process.

[0052] 6. The present application combines the type, pixel position and layer position of the defect, provides a more detailed and comprehensive defect description method, and improves the defect detection quality.

[0053] 7. The detection algorithm of the present application has self-learning ability, and the defect classifier can be continuously improved with the expansion of sample data, and further improves the detection accuracy.

[0054] 8. The present application can display the detection result in real time, has low operation difficulty, and is friendly in user interaction.

[0055] 9. The present application has high accuracy in detecting the defects of the flexible OLED panel, and reaches more than 98%. DETAILED DESCRIPTION

[0056] Fig. 1 is a schematic diagram of the device structure;

[0057] Fig. 2 is a detection flowchart of the device. DETAILED DESCRIPTION

[0058] The application discloses a flexible OLED panel defect detection device based on polarized light, and Figure 1 is a mechanical structure diagram of a system hardware part, which comprises a detector 1, an analyzer 2, a second 1 / 4 wave plate 3, a light source 4, a polarizer 5, a first 1 / 4 wave plate 6, a conveying belt 7, a scale disc 8, a motorized rotary table 9, a computer 10, a flexible OLED panel sample to be detected 11, a scale disc shaft support B 12 and a scale disc shaft support A 13. The device can be divided into:

[0059] A sample conveying device: comprising the conveying belt 7, which is used for conveying the flexible OLED panel sample to be detected 11 to a collection area and conveying the measured flexible OLED panel sample to a subsequent work station after detection;

[0060] A collection device: comprising the detector 1, the light source 4, wherein the light source is used for lighting on the surface of the measured flexible OLED panel sample, and the detector is used for acquiring the light intensity value and the power value of each point of the measured flexible OLED panel sample;

[0061] A polarization device: comprising the analyzer 2, the second 1 / 4 wave plate 3, the polarizer 5 and the first 1 / 4 wave plate 6, which are used for acquiring the polarization information of the measured flexible OLED panel sample;

[0062] An angle transformation device: comprising the motorized rotary table 9 and the scale disc 8, wherein the motorized rotary table is used for controlling the angle transformation of the polarization device, and the scale disc is used for controlling the angle transformation of the collection device;

[0063] The computer: is used for synchronously controlling the collection speed of the collection device and the movement speed of the sample conveying device, acquiring the light intensity value and the power value of each point of the measured flexible OLED panel sample at different time points through the collection device, controlling the angle transformation of the angle transformation device to acquire multiple sets of light intensity values and power values at different time points, realizing the calculation of different parameters, controlling the switching of the use state of the first 1 / 4 wave plate 6 and the second 1 / 4 wave plate 3 and the switching of the light source type, and realizing the acquisition of different parameters of the defect type and the defect level position;

[0064] The scale disc comprises the scale disc shaft support A 13 and the scale disc shaft support B 12, the scale disc shaft support A is fixed with the detector 1, the analyzer 2 and the second 1 / 4 wave plate 3 from top to bottom, the scale disc shaft support B is fixed with the light source 4, the polarizer 5 and the first 1 / 4 wave plate 6 from top to bottom, and the computer is connected with the rotary table 9, the conveying belt 7, the scale disc 8 and the light source 4 respectively.

[0065] The collection area is that the flexible OLED panel sample to be detected 11 is located in the visual range of the detector and the illumination range of the light source 4 at the same time.

[0066] The rotation of the angles of the polarizer 2, the second 1 / 4 wave plate 3, the polarizer 5 and the first 1 / 4 wave plate 6 in the device is realized by the rotation of the electric rotary table 9 located below, the rotation of the scale dial shaft support B realizes the rotation of the angle of the incident light, the rotation of the scale dial shaft support A realizes the rotation of the angle of the scattered light, and the rotation of the electric rotary table 9, the scale dial shaft support A 13 and the scale dial shaft support B 12 is realized by computer control.

[0067] The application further discloses a flexible OLED panel defect detection device based on polarized light and a detection method.

[0068] Step one: after the system is started, the light source 4 is switched to a white light source, the motion of the conveying belt is controlled until the standard sample is placed in the measurement range, and the Stokes vectors of the defect area and the non-defect area are obtained under multiple sets of experimental angles.

[0069] The multiple sets of experimental angles are that the incident light angle is in the range of 0° to 80° and every 10° is a set of selected incident light angles; and the PSG angle is in the range of 0° to 180° and every 10° is a set of selected PSG angles.

[0070] The step of calculating the Stokes vector is as follows:

[0071] 1) the light intensity values under six kinds of combinations of the PSA (0°, 0°), (0°, 45°), (45°, 45°), (90°, 90°), (0°, 135°) and (135°, 135°) are recorded, and the light intensity values of each pixel point obtained under the six sets of angle combination conditions are I1, I2, I3, I4, I5 and I6 respectively;

[0072] 2) the Stokes vector S = [A1, A2, A3, A4] of each pixel point is calculated by combining the light intensity values of each pixel point with the formula A1 = (I1 + I4), A2 = (I1 - I4), A3 = (I3 - I6) and A4 = (I5 - I2).

[0073] Step two: the Mueller matrix parameters of the defect area and the non-defect area are calculated through the Stokes vector, the optimal angle state is selected from the multiple angles, and the trained classifier is obtained by using the Mueller matrix parameter values under the optimal angle state for training the classifier.

[0074] The calculation of the Mueller matrix parameters is as follows: the incident light Stokes vector is S, the exit light Stokes vector is S', and the Mueller matrix parameter M of each pixel point can be calculated according to the formula S' = M·S, the M parameter is a four-row and four-column matrix, and the value of the first row and the first column is taken as the parameter value as the input data for training the classifier.

[0075] The selection step of the optimal angle state is as follows:

[0076] 1) Calculate the Mueller matrix parameters of each type of defect region and the Mueller matrix parameters of the defect-free region under each incident light angle and each PSG angle;

[0077] 2) Calculate the deviation percentage of each type of defect region compared to the defect-free region based on the Mueller matrix parameters of the defect-free region;

[0078] 3) Set the deviation percentage equal to 30% as the reference line for distinguishing between defect-free regions and defect regions, and 50% as the reference line for distinguishing between different defect regions, select the incident light angle and PSG angle that meet the distinction criteria, if there are multiple sets of angles that meet the criteria, select the set with the largest deviation as the optimal angle state, if there is only one set of angles that meet the criteria, select that set as the optimal angle state.

[0079] The training steps of the classifier are:

[0080] 1) Adjust the SVM support vector machine classifier to a classifier that needs to distinguish between the number of defect types (i.e., if there are N types of defects, adjust the SVM support vector machine to a (N+1) classifier);

[0081] 2) Perform multiple sets of repetitive experiments on the optimal angle state in the adjusted classifier, calculate the multiple sets of Mueller matrix parameter deviation values as input data sets for the training of the classifier, and save the trained classifier model for future use.

[0082] Step three: adjust the scale dial shaft holder B12, the polarizer 5 and the first 1 / 4 wave plate 6 to the optimal angle state, start the conveyor belt 7, control the conveyor belt to move until the flexible OLED panel sample to be measured is placed in the measurement range, control the motorized rotary stage 9 to obtain the Stokes vector of each pixel point of the flexible OLED panel sample to be measured;

[0083] The steps for calculating the Stokes vector are:

[0084] 1) Record the light intensity values of the PSA under six combinations of (0°, 0°), (0°, 45°), (45°, 45°), (90°, 90°), (0°, 135°), (135°, 135°), and record the light intensity values of each pixel point obtained under the six angle combinations as I1, I2, I3, I4, I5, I6, respectively;

[0085] 2) Calculate the Stokes vector S = [A1, A2, A3, A4] of each pixel point by combining the light intensity values of each pixel point with the formulas A1 = (I1 + I4), A2 = (I1 - I4), A3 = (I3 - I6), A4 = (I5 - I2).

[0086] Step four: calculate the Mueller matrix parameters of each pixel point of the flexible OLED panel sample to be tested by the Stokes vector of each pixel point, input the trained classifier for recognition and detection, and obtain the defect type and pixel point position;

[0087] The calculation of the Mueller matrix parameters is as follows: let the incident light Stokes vector be S and the outgoing light Stokes vector be S', the Mueller matrix parameters M of each pixel point can be calculated according to the formula S' = M·S, the M parameter is a four-row four-column matrix, and the deviation of the value of the first row and the first column from the parameter value of the defect-free area under the same angle state of the standard sample is taken as the input data of the detection classifier.

[0088] Step five: control the light source 4 to switch to a laser, the first 1 / 4 wave plate 6 and the second 1 / 4 wave plate 3 to switch to a non-use state, the scale disc shaft support B12 to adjust the incident light angle to a fixed value, and the scale disc shaft support A13 to adjust the scattering light receiving angle within an angle range, and obtain the power value measured by the detector 1;

[0089] The fixed value of the incident light angle is 60°, and the scattering light receiving angle range is 36°-72°.

[0090] Step six: calculate the change trend of the polarization factor of each pixel point of the flexible OLED panel sample to be tested by the power value, analyze whether the defect position belongs to the surface, the subsurface or the rough surface, and obtain the flexible OLED panel level position of the defect;

[0091] The change trend calculation steps of the polarization factor are as follows:

[0092] 1) convert the power value into BRDF pp ;

[0093] 2) draw the BRDF pp change curve to obtain the trend of the BRDF pp change with the scattering light angle.

[0094] The BRDF pp expression calculation and the scattering light angle change classification are discussed as follows.

[0095] 1) if the defect is above the surface, the Fresnel reflection coefficients of s light and p light are calculated according to the following formula:

[0096] The polarization coefficient calculation formula is:

[0097] The polarization bidirectional reflection distribution function calculation formula is:

[0098] 2) if the defect is above the surface, the polarization coefficient calculation formula is:

[0099] The formula for calculating the polarized bidirectional reflectance distribution function is:

[0100] 3) If the defect is on the micro-rough surface, the formula for calculating the polarization coefficient is:

[0101] The formula for calculating the polarized bidirectional reflectance distribution function is:

[0102] Comprehensive comparison of the polarization coefficients q ss , q sp , q ps , q pp of the BRDFs of the surface, subsurface, and micro-rough surface ss , q sp , q ps Each of the three terms has a common function term, so the three scattering mechanisms will be superimposed during scattering measurement, and the three scattering mechanisms cannot be effectively separated. pp There is no separate common term, and the expressions of the three are obviously different. The relationship between the BRDF pp and the scattering angle is different, which can effectively distinguish the defect positions in the three cases.

[0103] The judgment basis for the defect position belonging to the surface, subsurface, or micro-rough surface is:

[0104] 1) If the BRDF pp increases with the increase of the scattering azimuth angle, it indicates that the defect at this position is on the surface;

[0105] 2) If the BRDF pp has a minimum value near the scattering azimuth angle of 36°, and the BRDF part changes with the scattering azimuth angle in other ranges, it indicates that the defect at this position is on the micro-rough surface;

[0106] 3) If the BRDF pp has a minimum value near the scattering azimuth angle of 72°, and the BRDF part changes with the scattering azimuth angle in other ranges, it indicates that the defect at this position is in the subsurface.

[0107] Step seven: fuse the flexible OLED panel hierarchical position as supplementary information into the defect type and pixel position conclusion, and obtain the comprehensive result of defect detection as the conclusion output.

[0108] Finally, it should be noted that the above enumeration is only a specific embodiment of the present application. Obviously, the present application is not limited to the above embodiment, and there can be many variations. All variations that can be directly derived or inferred from the disclosure of the present application by those of ordinary skill in the art should be considered within the scope of the present application.

Claims

1. A polarization-based flexible OLED panel defect detection device, characterized in that, The device comprises: a sample conveying device comprising a conveying belt for conveying the flexible OLED panel sample to be measured to a collection area and conveying the measured flexible OLED panel sample to a subsequent station after detection; a collection device comprising a detector and a light source, wherein the light source is used for lighting the surface of the measured flexible OLED panel sample, and the detector is used for acquiring the light intensity value and power value of each point of the measured flexible OLED panel sample; a polarization device comprising an analyzer, a second 1 / 4 wave plate, a polarizer and a first 1 / 4 wave plate, which is used for acquiring the polarization information of the measured flexible OLED panel sample; an angle conversion device comprising a motorized rotary table and a dial, wherein the motorized rotary table is used for controlling the angle conversion of the polarization device, and the dial is used for controlling the angle conversion of the collection device; a computer, which is used for synchronously controlling the collection speed of the collection device and the movement speed of the sample conveying device, acquiring the light intensity value and power value of each point of the measured flexible OLED panel sample at different time points, controlling the angle conversion of the angle conversion device, and acquiring multiple sets of light intensity value and power value at different time points to realize the calculation of different parameters, controlling the switching of the use state of the first 1 / 4 wave plate and the second 1 / 4 wave plate and the switching of the light source type, and realizing the acquisition of different parameters of the defect type and the defect level position; the dial comprises a dial shaft support A and a dial shaft support B, the dial shaft support A is fixed with the detector, the analyzer and the second 1 / 4 wave plate from top to bottom, the dial shaft support B is fixed with the light source, the polarizer and the first 1 / 4 wave plate from top to bottom, and the computer is connected with the motorized rotary table, the conveying belt, the dial and the light source. the collection area is located within the visual range of the detector and the illumination range of the light source. 2.The polarization-based flexible OLED panel defect detection device of claim 1, wherein, the rotation of the angles of the analyzer, the second 1 / 4 wave plate, the polarizer and the first 1 / 4 wave plate in the device is realized by the rotation of the motorized rotary table below, the rotation of the incident light angle is realized by the rotation of the dial shaft support B, the rotation of the scattered light angle is realized by the rotation of the dial shaft support A, and the rotation of the motorized rotary table, the dial shaft support A and the dial shaft support B is realized by the control of the computer. 3.A method for defect detection of a flexible OLED panel based on polarized light, characterized in that, The device comprises the following steps: 1) switching the light source to a white light source, controlling the movement of the conveying belt until the standard sample is placed in the measurement range, and obtaining the Stokes vector of the defect area and the non-defect area at multiple experimental angles; 2) calculating the Mueller matrix parameters of the defect area and the non-defect area from the Stokes vector, selecting the optimal angle state from the multiple angles, training the classifier with the Mueller matrix parameter value at the optimal angle state, and obtaining the trained classifier; 3) adjusting the angles of the dial shaft support B, the polarizer and the first 1 / 4 wave plate to the optimal angle state, starting the conveying belt, controlling the movement of the conveying belt until the flexible OLED panel sample to be measured is placed in the measurement range, and controlling the motorized rotary table to obtain the Stokes vector of each pixel point of the measured flexible OLED panel sample. 4) Calculate the Mueller matrix parameters of each pixel point of the flexible OLED panel sample to be measured by the Stokes vector of each pixel point, input the trained classifier for recognition and detection to obtain the defect type and pixel point position; 5) Control the light source to switch to a laser, and the first 1 / 4 wave plate and the second 1 / 4 wave plate to be in a non-use state, adjust the incident light angle to a fixed value by the dial shaft support B, adjust the scattering light receiving angle within the angle range by the dial shaft support A, and obtain the power value measured by the detector; 6) Calculate the change trend of the polarization factor of each pixel point of the flexible OLED panel sample to be measured by the power value, analyze the defect position belonging to the surface, subsurface or rough surface, and obtain the flexible OLED panel level position of the defect; 7) Fuse the flexible OLED panel level position as supplementary information into the defect type and pixel point position result, and obtain the comprehensive result of the defect detection as the conclusion output. 4.The polarization-based flexible OLED panel defect detection method of claim 3, wherein, The multiple sets of experimental angles in step 1) are that the incident light angle is in the range of 0° to 80°, and every 10° is a set of selected incident light angles; and the PSG angle is in the range of 0° to 180°, and every 10° is a set of selected PSG angles. The fixed value of the incident light angle in step 5) is 60°, and the scattering light receiving angle range is 36°-72°. 5.The polarization-based flexible OLED panel defect detection method of claim 3, wherein, The steps for calculating the Stokes vector in steps 1) and 3) are: 1) Record the light intensity values under six combinations of PSA at (0°, 0°), (0°, 45°), (45°, 45°), (90°, 90°), (0°, 135°) and (135°, 135°), and denote the light intensity values of each pixel point obtained under the six angle combination conditions as I1, I2, I3, I4, I5 and I6, respectively; 2) Calculate the Stokes vector S = [A1, A2, A3, A4] of each pixel point by the light intensity values of each pixel point and the formulae A1 = (I1 + I4), A2 = (I1 - I4), A3 = (I3 - I6) and A4 = (I5 - I2).

6. The polarized light based flexible OLED panel defect detection method of claim 3 or 4, wherein, The calculation of the Mueller matrix parameters in step 2) is that the incident light Stokes vector is denoted as S, the outgoing light Stokes vector is denoted as S', and the Mueller matrix parameters M of each pixel point can be calculated according to the formula S' = M·S. The M parameter is a four-row and four-column matrix, and the value of the first row and the first column is taken as the parameter value as the input data of the training classifier. The calculation of the Mueller matrix parameters in step 4) is that the incident light Stokes vector is denoted as S, the outgoing light Stokes vector is denoted as S', and the Mueller matrix parameters M of each pixel point can be calculated according to the formula S' = M·S. The M parameter is a four-row and four-column matrix, and the value of the first row and the first column is taken to calculate the deviation of the parameter value of the defect-free area under the same angle condition as the standard sample as the input data of the detection classifier.

7. The polarized light based flexible OLED panel defect detection method of claim 6, wherein, The selection steps of the optimal angle state in step 2) are: 1) Calculate the Mueller matrix parameters of each type of defect area and the Mueller matrix parameters of the defect-free area under each incident light angle and each PSG angle; 2) Calculate the deviation percentage of each type of defect area compared to the non-defect area based on the Mueller matrix parameters of the non-defect area as the standard; 3) Set the deviation percentage equal to 30% as the reference line for distinguishing between the defect area and the non-defect area, and equal to 50% as the reference line for distinguishing between different defect areas, select the incident light angle and the PSG angle that meet the distinguishing criteria, if there are multiple groups of angles that meet the criteria, select the group with the largest deviation as the optimal angle state, if there is only one group of angles that meet the criteria, select this group as the optimal angle state.

8. The polarized light based flexible OLED panel defect detection method of claim 3 or 4 or 5 or 7, wherein, The training step of the classifier in step 2) is: 1) Adjust the SVM support vector machine classifier to a classifier that needs to distinguish (N+1) types of defects (i.e., if there are N types of defects, adjust the SVM support vector machine to (N+1) classifiers); 2) Perform multiple sets of repetitive experiments on the optimal angle state in the adjusted classifier, calculate the multiple sets of Mueller matrix parameter deviation values as input data sets for the training of the classifier, and save the trained classifier model for future use. 9.The polarization-based flexible OLED panel defect detection method of claim 8, wherein, The step 6) is the change trend calculation step of the polarization factor: 1) Convert power values to BRDF pp ; 2) plot BRDF pp variation curve, resulting BRDF pp trend of variation with angle of scattered light.

10. The polarized light based flexible OLED panel defect detection method of claim 9, wherein, The judgment basis for whether the defect position belongs to the surface, subsurface or micro-rough surface in step 6) is: 1) if BRDF pp increases with increasing scattering azimuth angle, indicating a defect above the surface; 2) if BRDF pp The minimum value appears near the scattering azimuth angle of 36°, and BRDF pp The change with the scattering azimuth angle indicates that the defect is on the micro-rough surface. 3) if BRDF pp The minimum value appears near the scattering azimuth angle of 72°, and BRDF pp The change with the scattering azimuth angle indicates that the defect is in the subsurface.

Images