Display screen defect detection method based on machine vision

Abstract

The invention relates to a display screen defect detection method based on machine vision, and the method comprises the following steps: obtaining an order area of a to-be-detected display, matching and determining a detection correction item, and correcting a defect detection standard of the to-be-detected display; collecting a five-color picture set of the to-be-detected display; preprocessing the five-color picture set to obtain a five-color picture set to be detected; and performing defect detection on the to-be-detected five-color picture set through a preset visual detection model based on the corrected defect detection standard, and outputting a defect detection result. The method has the effect of effectively improving the detection efficiency and the detection precision.

Description

Technical Field

[0001] The present application relates to the field of visual inspection, and in particular to a display screen defect detection method based on machine vision. Background Art

[0002] With the continuous development of science and technology, display screens are increasingly used in various fields. However, various defects may occur in the display screen during the production process, such as bright spots, dark spots, line defects, etc. These defects not only affect the appearance quality of the display screen, but also may affect its performance and service life. Therefore, it is very necessary to perform defect detection on the display screen. The traditional display screen defect detection method mainly relies on manual detection, which is not only inefficient and costly, but also the detection results are easily affected by human factors.

[0003] Existing display manufacturers often have businesses all over the country and even export to overseas. However, the sensitivity of the human eye to RGB (red, green, and blue) varies in different latitudes. The traditional machine vision inspection methods currently used by companies have received different levels of feedback in different regions. Users in high-latitude areas often report that some blue bright spot defects are more obvious during use, but are not discovered during inspection; while product inspectors in low-latitude areas often feel that some displays with slight color differences in green and blue cannot be seen by the naked eye during re-inspection, but are judged as defects by the inspection system, and feedback indicates that there are problems with the inspection standards.

[0004] In order to solve these problems, a display screen defect detection method based on machine vision is needed. Summary of the invention

[0005] In order to solve the above problems, the present application provides a display screen defect detection method based on machine vision.

[0006] In a first aspect, the present application provides a display screen defect detection method based on machine vision, which adopts the following technical solution:

[0007] A display screen defect detection method based on machine vision comprises the following steps:

[0008] Obtain the region to which the order of the display to be inspected belongs, match and determine the inspection correction items, and correct the defect inspection standard of the display to be inspected;

[0009] Collecting a five-color picture set of the display to be detected, wherein the five-color picture set includes a black picture, a white picture, a red picture, a green picture and a blue picture of the display to be detected;

[0010] Preprocessing the five-color picture set to obtain the five-color picture set to be detected;

[0011] Based on the revised defect detection standard, defect detection is performed on the set of five-color images to be inspected through a preset visual inspection model, and the defect detection results are output.

[0012] Preferably, the obtaining of the region to which the order of the display to be inspected belongs, matching and determining the inspection correction item, and correcting the defect inspection standard of the display to be inspected specifically include the following steps:

[0013] Obtain order information of the display to be tested, and determine the region to which the order belongs; the order information includes basic information of the display to be tested, information on the number of displays, and the region to which the order belongs;

[0014] Obtaining defect detection standards of the display to be inspected based on the order information;

[0015] Determine the regional characteristic label based on the region to which the order belongs;

[0016] Determine the detection correction item through matching with a preset correction matching model according to the regional characteristic label; the correction matching model is a neural network model, which is obtained through deep learning training through human eye visual characteristic data of different latitudes, display sample data, environmental care data and user color preference questionnaire;

[0017] The defect detection standard of the display to be inspected is modified according to the detection correction items.

[0018] Preferably, the step of determining the regional characteristic tag according to the region to which the order belongs includes the following steps:

[0019] Obtain the latitude range of the area to which the order belongs, and generate a latitude label of the display to be detected, wherein the latitude label includes a latitude type and a latitude representative value, and the latitude type includes high latitude, middle latitude, and low latitude;

[0020] Connecting to the Internet public database to obtain cultural characteristics information of the area to which the order belongs, and generating a cultural label for the area to which the order belongs; the cultural label includes a country label, an ethnic proportion label, and a cultural label;

[0021] The latitude label and humanistic label of the area to which the order belongs are packaged to generate regional characteristic labels.

[0022] Preferably, the step of obtaining the latitude range of the area to which the order belongs and generating the latitude label of the display to be detected specifically comprises the following steps:

[0023] Get the latitude range of the area to which the order belongs, and determine the maximum and minimum latitude of the area to which the order belongs;

[0024] Determine whether the maximum latitude of the area to which the order belongs is within the high latitude interval, where the high latitude interval is [60, 90];

[0025] If it does, the region to which the order belongs is determined to be a high-latitude region, and a latitude label for the region to which the order belongs is generated, where the latitude type is high latitude and the latitude representative value is the maximum latitude of the region to which the order belongs;

[0026] If not, determine whether the minimum latitude of the area to which the order belongs is within a low latitude interval, where the low latitude interval is [0, 30];

[0027] If it does, the region to which the order belongs is determined to be a low-latitude region, and a latitude label for the region to which the order belongs is generated, where the latitude type is low latitude and the latitude representative value is the minimum latitude value of the region to which the order belongs;

[0028] If not, the area to which the order belongs is determined to be a mid-latitude area, and a latitude label for the area to which the order belongs is generated, where the latitude type is low latitude and the latitude representative value is the median latitude of the area to which the order belongs.

[0029] Preferably, the correction of the defect detection standard of the display to be inspected according to the detection correction item also includes: classifying the corrected defect detection standards based on the detection correction item to obtain a standard improvement set and a standard reduction set, and sorting each defect detection item in order of standard improvement, standard unchanged and standard reduction to generate a detection process.

[0030] Preferably, the collecting of the five-color picture set of the display to be detected specifically includes the following steps:

[0031] Obtain natural lighting information of the area to which the order of the display to be tested belongs, and simulate the natural lighting environment of the area to which the order belongs through a preset fill light source in a closed environment;

[0032] The screen images of the display are switched to black, white, red, green and blue in sequence, and the five-color images of the display screen area are collected by a high-definition camera to generate a daytime use atlas;

[0033] Obtain indoor lighting information of the area to which the order of the display to be detected belongs, and simulate the indoor lighting environment of the area to which the order belongs through a preset fill light source in a closed environment;

[0034] The screen images of the display are switched to black, white, red, green and blue in sequence, and a high-definition camera is used to capture the five-color images of the display screen area to generate a night use atlas;

[0035] The daytime usage atlas and the nighttime usage atlas are packaged to generate a five-color picture set of the display to be tested.

[0036] Preferably, the defect detection of the five-color picture set to be detected is performed by using a preset visual detection model based on the revised defect detection standard, and outputting the defect detection result specifically includes the following steps:

[0037] Input the corrected defect detection standard into the preset visual inspection model to set the inspection standard;

[0038] Inputting the five-color images in the five-color image set to be detected into a preset visual detection model, wherein the visual detection model is a convolutional neural network model obtained by deep learning training of standard five-color image sample data of the display;

[0039] The visual inspection model roughly searches for defects in the five-color image to obtain the ROI area where defects may exist in the five-color image;

[0040] The visual inspection model performs defect detection on the ROI area of ​​the four-color image according to the defect detection standard and outputs the defect detection results.

[0041] Preferably, the preprocessing of the five-color picture set to obtain the five-color picture set to be detected specifically comprises the following steps:

[0042] Filter each color image in the five-color image set to remove texture background and noise points in each color image;

[0043] Performing sharpening processing on each color image in the five-color image set after filtering, so as to enhance the sharpness and clarity of each color image in the five-color image set;

[0044] Crop each color picture in the five-color picture set after the sharpening process, remove the environmental background and the display frame of the display to be detected in each color picture, and only keep the picture of the screen area of ​​the display to be detected;

[0045] The cropped five-color picture set is normalized to obtain a five-color picture set to be detected.

[0046] In a second aspect, a computer-readable storage medium is provided, characterized in that it stores a computer program that can be loaded by a processor and execute any of the above methods.

[0047] In summary, the present application includes at least one of the following beneficial technical effects:

[0048] 1. Based on the visual characteristics of the human eye, according to the latitude information of the area required by the order of the display to be inspected and the cultural characteristics, the defect detection standards of the display to be inspected are revised to realize intelligent customized detection standards, which is helpful to improve the detection accuracy, make the display color defect detection more accurate, and the defect detection of brightness and contrast more in line with the actual use situation, fit the user's visual experience, improve the user experience, help avoid overly strict or loose detection standards, reduce unnecessary rework or scrapping caused by misjudgment, make more effective use of detection resources, and effectively improve the detection efficiency and detection accuracy;

[0049] 2. The visual inspection model selects ROI areas where defects may exist based on ROI architecture rules, and finally performs detailed feature collection and defect inspection on the obtained ROI areas. This can greatly reduce the defect detection workload of the visual inspection model, reduce the data processing load, and improve the accuracy and efficiency of conventional defect detection. BRIEF DESCRIPTION OF THE DRAWINGS

[0050] Figure 1 is a method flow chart of a display screen defect detection method based on machine vision in an embodiment of the present application;

[0051] Figure 2 is a flow chart of a method for correcting a defect detection standard of a display to be detected in an embodiment of the present application;

[0052] Figure 3 is a flow chart of a method for matching and determining regional characteristic tags of the region to which an order belongs in an embodiment of the present application;

[0053] Figure 4 is a flow chart of a method for generating a latitude label of a display to be detected in an embodiment of the present application;

[0054] Figure 5 is a flow chart of a method for collecting a set of five-color images of a display to be detected in an embodiment of the present application;

[0055] Figure 6 is a flow chart of a method for preprocessing a five-color picture set in an embodiment of the present application;

[0056] Figure 7 It is a flow chart of a method for performing defect detection on a set of five-color images to be detected in an embodiment of the present application. DETAILED DESCRIPTION

[0057] The following is combined with Figure 1-7 This application is described in further detail.

[0058] There are three different types of cones on the retina of the human eye, which are sensitive to red, green, and blue wavelengths of light. The relative stimulation of these three cones determines the human eye's perception of different colors. Generally speaking, the human eye is most sensitive to green, followed by red, and relatively less sensitive to blue. The sensitivity of the human eye to different colors varies depending on the region or latitude. The light in high-latitude areas is relatively weak, the angle of sunlight is low, and the sky usually appears lighter blue. Due to the low light intensity, the human eye is relatively sensitive to brightness. In this case, the human eye may be more sensitive to blue in the RGB tricolor because blue is easier to detect in low-brightness environments. The sunlight in low-latitude areas is strong, the angle of light is high, and the sky usually appears deeper blue. Strong light makes the human eye relatively less sensitive to brightness. In this case, the human eye may be relatively less sensitive to green and blue in the RGB tricolor because these two colors are easily confused with white in high-brightness environments.

[0059] In addition, the cultural background and aesthetic concepts of different latitudes will also affect people's tolerance and attention to display screen defects. For example, people in our country are highly sensitive to red. Taking Thailand as an example, Thailand's cultural art forms such as traditional temple architecture and dance costumes are extremely bright in color. In Thai temple architecture, highly saturated red, yellow, green and other colors are often used. These color combinations are bright and high in contrast. Under the influence of such a cultural environment, people's preference for display screen colors also tends to be bright and high in contrast.

[0060] The present application embodiment discloses a display screen defect detection method based on machine vision. Figure 1 , a display screen defect detection method based on machine vision, comprising the following steps:

[0061] S1. Standard correction: obtain the region to which the order of the display to be tested belongs, match and determine the test correction items, and correct the defect detection standard of the display to be tested;

[0062] S2. Collecting pictures: collecting a set of five-color pictures of the display to be detected, wherein the set of five-color pictures includes a black picture, a white picture, a red picture, a green picture and a blue picture of the display to be detected;

[0063] S3, picture preprocessing: preprocessing the five-color picture set to obtain the five-color picture set to be detected;

[0064] S4. Defect detection: Based on the corrected defect detection standard, the five-color screen set to be inspected is subjected to defect detection through the preset visual inspection model, and the defect detection results are output. Through the above steps, starting from the visual characteristics of the human eye, based on the latitude information of the area required for the order of the display to be inspected and the cultural characteristics, the defect detection standard of the display to be inspected is corrected, and intelligent customized inspection standards are realized, which helps to improve the inspection accuracy, makes the color defect detection of the display more accurate, and the defect detection of brightness and contrast more in line with actual usage, fits the user's visual perception, improves the user experience, helps to avoid overly strict or loose inspection standards, reduces unnecessary rework or scrapping due to misjudgment, and makes more effective use of inspection resources (such as time, manpower, equipment, etc.), so as to effectively improve the inspection efficiency and inspection accuracy.

[0065] Reference Figure 2 The obtaining of the order area of ​​the display to be inspected, matching and determining the inspection correction items, and correcting the defect inspection standard of the display to be inspected specifically include the following steps:

[0066] A1. Determine the region: obtain the order information of the display to be tested and determine the region to which the order belongs; the order information includes basic information of the display to be tested, information on the number of displays, and the region to which the order belongs;

[0067] A2. Acquisition standard: Acquisition of defect detection standard of the display to be inspected based on order information;

[0068] A3. Matching tags: Determine regional characteristic tags based on the region to which the order belongs;

[0069] A4. Determine detection correction items: Determine detection correction items by matching a preset correction matching model according to the regional characteristic labels; the correction matching model is a neural network model, which is obtained by deep learning training through human eye visual characteristic data of different latitudes, display sample data, environmental care data and user color preference questionnaire; it should be noted that the specific training steps of the neural network model are existing technologies and will not be repeated here;

[0070] A5. Standard correction: Correct the defect detection standard of the display to be inspected according to the detection correction items. Through the above steps, the regional characteristic tags are matched according to the region to which the order belongs, and then the detection correction items are matched according to these characteristic tags by correcting the matching model, and the defect detection standard of the display to be inspected is corrected to realize intelligent customization of the detection standard, so that the detection standard is more in line with the actual detection needs of the region where the order is located, which helps to avoid overly strict or loose detection standards, reduce unnecessary rework or scrapping due to misjudgment, and effectively improve the detection efficiency and detection accuracy.

[0071] Reference Figure 3 The step of determining the regional characteristic tag according to the region to which the order belongs includes the following steps:

[0072] B1. Generate latitude label: obtain the latitude range of the area to which the order belongs, and generate the latitude label of the display to be tested. The latitude label includes the latitude type and the latitude representative value. The latitude type includes high latitude, middle latitude and low latitude;

[0073] B2. Generate humanistic tags: connect to the Internet public database to obtain the humanistic characteristics information of the area to which the order belongs, and generate humanistic tags for the area to which the order belongs; the humanistic tags include country tags, ethnic proportion tags and cultural tags;

[0074] B3. Generate regional characteristic labels by packaging: Pack the latitude label and humanistic label of the region to which the order belongs to generate regional characteristic labels. For the display to be tested, the latitude factors (physiological factors) and humanistic factors (cultural and psychological factors) of the region to which the order belongs are comprehensively considered to formulate regional characteristic labels for the region to which the order belongs, which is conducive to realizing intelligent customized testing standards.

[0075] Reference Figure 4 The step of obtaining the latitude range of the area to which the order belongs and generating the latitude label of the display to be detected specifically includes the following steps:

[0076] C1. Determine the latitude range: obtain the latitude range of the area to which the order belongs, and determine the maximum and minimum latitude of the area to which the order belongs;

[0077] C2. Determine whether the maximum latitude of the area to which the order belongs is within the high latitude interval, where the high latitude interval is [60, 90];

[0078] C3. The region to which the order belongs is a high-latitude region: If so, the region to which the order belongs is determined to be a high-latitude region, and a latitude label of the region to which the order belongs is generated, where the latitude type is high latitude and the latitude representative value is the maximum latitude of the region to which the order belongs;

[0079] C4. If not, determine whether the minimum latitude of the area to which the order belongs is within a low latitude interval, where the low latitude interval is [0, 30];

[0080] C5. The region to which the order belongs is a low-latitude region: If so, the region to which the order belongs is determined to be a low-latitude region, and a latitude label of the region to which the order belongs is generated, where the latitude type is low latitude and the latitude representative value is the minimum latitude value of the region to which the order belongs;

[0081] C6. The region to which the order belongs is a mid-latitude region: If not, the region to which the order belongs is determined to be a mid-latitude region, and a latitude label for the region to which the order belongs is generated, where the latitude type is low latitude, and the latitude representative value is the median latitude of the region to which the order belongs.

[0082] For example, if the order area is Thailand, and the latitude range of Thailand is 10-20° north latitude, the latitude label is the latitude type low latitude, and the latitude representative value is 10. Similarly, the humanistic labels include the national label Thailand, the ethnic proportion label 40% Thai, 35% Lao, the cultural label is the state religion Buddhism, and the color culture. Based on these regional characteristic labels, the detection correction items output by the correction matching model are: 1. Lower the overall brightness uniformity detection standard. For example, the deviation of the brightness uniformity of the standard defect detection may be required to be within 5%, while in Thailand, it can be relaxed to within 8%; (latitude factor) 2. Lower the blue channel detection standard, and increase the allowable number of blue bright spots and dark spots. For example, in the conventional detection standard, the number of blue bright spots allowed on the screen is 3 1. The detection threshold of blue bright spots should be within 3, and it can be increased to within 5 in Thailand. At the same time, the detection threshold of blue dark spots should be increased and the detection threshold of blue bright spots should be lowered; (latitude factor); 3. Improve the detection standards of red and green channels, because the local human eyes are more sensitive to these two colors (religion). For example, the color deviation (Delta-E value) of red may be allowed to be within 3 in the conventional detection standard, but it is required to be within 2 in Thailand; and the detection thresholds of red and green bright spots should be lowered, and the detection thresholds of red and green dark spots should be increased; (cultural factor) 4. Improve the detection standards of color saturation and contrast. For example, in the conventional detection standards, if the color saturation is lower than 60 or the contrast is lower than 250:1, it will be marked for subsequent adjustment. In Thailand, the color saturation is set to be lower than 70 or the contrast is lower than 300:1 for marking. (Cultural factor)

[0083] In addition, after the defect detection standard of the display to be inspected is corrected according to the detection correction item, the corrected defect detection standards are classified based on the detection correction item to obtain a standard improvement set and a standard reduction set, and each defect detection item is sorted in order according to standard improvement, standard unchanged, and standard reduction to generate a detection process. By sorting each defect detection item and giving priority to the detection items with improved detection standards, defective products can be quickly screened out, the detection workload can be reduced, and the detection efficiency can be improved.

[0084] Reference Figure 5 The step of collecting the five-color picture set of the display to be detected specifically includes the following steps:

[0085] D1. Simulate natural lighting environment: obtain the natural lighting information of the order area of ​​the display to be tested, and simulate the natural lighting environment of the order area through a preset fill light source in a closed environment;

[0086] D2. Generate a daytime use atlas: Switch the screen image of the display to black, white, red, green and blue images in sequence, and use a high-definition camera to collect five-color images of the display screen area to generate a daytime use atlas;

[0087] D3. Simulate indoor lighting environment: obtain indoor lighting information of the area to which the order of the display to be tested belongs, and simulate the indoor lighting environment of the area to which the order belongs through a preset fill light source in a closed environment;

[0088] D4. Generate a night use atlas: Switch the screen image of the display to black, white, red, green and blue images in sequence, and use a high-definition camera to collect five-color images of the display screen area to generate a night use atlas;

[0089] D5. Packaging and generating a five-color picture set: Packing the daytime use atlas and the nighttime use atlas to generate a five-color picture set of the display to be tested. Based on the normal usage scenarios of the area where the display belongs, collecting the five-color pictures of the display helps to improve the accuracy of defect detection.

[0090] Reference Figure 6 The preprocessing of the five-color picture set to obtain the five-color picture set to be detected specifically includes the following steps:

[0091] E1. Image filtering: Filter each color image in the five-color image set to remove texture background and noise points in each color image;

[0092] E2. Image sharpening: sharpening each color image in the five-color image set after filtering, to enhance the sharpness and clarity of each color image in the five-color image set;

[0093] E3. Image cropping: cropping each color image in the five-color image set after sharpening processing, removing the environmental background and display frame of the display to be detected in each color image, and only retaining the image of the screen area of ​​the display to be detected;

[0094] E4. Normalization: Normalize the cropped five-color image set to obtain the five-color image set to be detected. The collected five-color images are pre-processed through the above steps to remove the texture background of the five-color images while enhancing the sharpness and clarity of the images, which helps to improve the accuracy of defect detection.

[0095] Reference Figure 7 The method of performing defect detection on the five-color picture set to be detected by using a preset visual detection model based on the revised defect detection standard and outputting the defect detection result specifically includes the following steps:

[0096] F1. Inspection standard setting: Input the corrected defect inspection standard into the preset visual inspection model to set the inspection standard;

[0097] F2. Picture input: input the five-color pictures in the five-color picture set to be detected into a preset visual detection model, where the visual detection model is a convolutional neural network model obtained through deep learning training of standard five-color picture sample data of the display;

[0098] F3. Find ROI area: The visual inspection model roughly searches for defects in the five-color image to obtain the ROI area where defects may exist in the five-color image;

[0099] F4. Defect detection: The visual inspection model performs defect detection on the ROI area of ​​the four-color image according to the defect detection standard and outputs the defect detection results. The visual inspection model selects the ROI area where defects may exist based on the ROI architecture rules, and finally performs detailed feature collection and defect inspection on the obtained ROI area, which can greatly reduce the defect detection workload of the visual inspection model, reduce the data processing load, and improve the accuracy and efficiency of conventional defect detection.

[0100] The embodiment of the present application also discloses a computer-readable storage medium, which stores a computer program that can be loaded by a processor and executed as in the above method. The computer-readable storage medium includes, for example: a USB flash drive, a mobile hard disk, a read-only memory (ROM), a random access memory (RAM), a magnetic disk or an optical disk, and other media that can store program codes.

[0101] The above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit the scope of protection of the invention. Obviously, the described embodiments are only some embodiments of the present invention, rather than all embodiments. Based on these embodiments, all other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of the present invention. Although the present invention has been described in detail with reference to the above embodiments, ordinary technicians in this field can still combine, add, delete or make other adjustments to the features in the various embodiments of the present invention according to the circumstances without conflict, without making creative work, so as to obtain different other technical solutions that do not deviate from the concept of the present invention in essence, and these technical solutions also belong to the scope of protection of the present invention.

Claims

1. A display screen defect detection method based on machine vision, characterized in that: The following steps are involved: Obtain the region to which the order of the display to be inspected belongs, match and determine the inspection correction items, and correct the defect inspection standard of the display to be inspected; Collecting a five-color picture set of the display to be detected, wherein the five-color picture set includes a black picture, a white picture, a red picture, a green picture and a blue picture of the display to be detected; Preprocessing the five-color picture set to obtain the five-color picture set to be detected; Based on the revised defect detection standard, defect detection is performed on the set of five-color images to be inspected through a preset visual inspection model, and the defect detection results are output.

2. The display screen defect detection method based on machine vision according to claim 1, characterized in that: The steps of obtaining the region to which the order of the display to be inspected belongs, matching and determining the inspection correction items, and correcting the defect inspection standard of the display to be inspected specifically include the following steps: Obtain order information of the display to be tested, and determine the region to which the order belongs; the order information includes basic information of the display to be tested, information on the number of displays, and the region to which the order belongs; Obtaining defect detection standards of the display to be inspected based on the order information; Determine the regional characteristic label based on the region to which the order belongs; Determine the detection correction item through matching with a preset correction matching model according to the regional characteristic label; the correction matching model is a neural network model, which is obtained through deep learning training through human eye visual characteristic data of different latitudes, display sample data, environmental care data and user color preference questionnaire; The defect detection standard of the display to be inspected is modified according to the detection correction items.

3. The display screen defect detection method based on machine vision according to claim 2, characterized in that: The method of determining the regional characteristic tag according to the region to which the order belongs specifically includes the following steps: Obtain the latitude range of the area to which the order belongs, and generate a latitude label of the display to be detected, wherein the latitude label includes a latitude type and a latitude representative value, and the latitude type includes high latitude, middle latitude, and low latitude; Connecting to the Internet public database to obtain cultural characteristics information of the area to which the order belongs, and generating a cultural label for the area to which the order belongs; the cultural label includes a country label, an ethnic proportion label, and a cultural label; The latitude label and humanistic label of the area to which the order belongs are packaged to generate regional characteristic labels.

4. The display screen defect detection method based on machine vision according to claim 3, characterized in that: The step of obtaining the latitude range of the area to which the order belongs and generating the latitude label of the display to be detected specifically includes the following steps: Get the latitude range of the area to which the order belongs, and determine the maximum and minimum latitude of the area to which the order belongs; Determine whether the maximum latitude of the area to which the order belongs is within the high latitude interval, where the high latitude interval is [60, 90]; If it does, the region to which the order belongs is determined to be a high-latitude region, and a latitude label for the region to which the order belongs is generated, where the latitude type is high latitude and the latitude representative value is the maximum latitude of the region to which the order belongs; If not, determine whether the minimum latitude of the area to which the order belongs is within a low-latitude interval, where the low-latitude interval is [0, 30]; If it does, the region to which the order belongs is determined to be a low-latitude region, and a latitude label for the region to which the order belongs is generated, where the latitude type is low latitude and the latitude representative value is the minimum latitude value of the region to which the order belongs; If not, the area to which the order belongs is determined to be a mid-latitude area, and a latitude label for the area to which the order belongs is generated, where the latitude type is low latitude and the latitude representative value is the median latitude of the area to which the order belongs.

5. The display screen defect detection method based on machine vision according to claim 4, characterized in that: The method of correcting the defect detection standard of the display to be inspected according to the detection correction item also includes: classifying the corrected defect detection standards based on the detection correction item to obtain a standard improvement set and a standard reduction set, and sorting the defect detection items in order of standard improvement, standard unchanged and standard reduction to generate a detection process.

6. The display screen defect detection method based on machine vision according to claim 1, characterized in that: The acquisition of the five-color picture set of the display to be detected specifically includes the following steps: Obtain natural lighting information of the area to which the order of the display to be tested belongs, and simulate the natural lighting environment of the area to which the order belongs through a preset fill light source in a closed environment; The screen images of the display are switched to black, white, red, green and blue in sequence, and the five-color images of the display screen area are collected by a high-definition camera to generate a daytime use atlas; Acquire the indoor lighting information of the area to which the order of the display to be detected belongs, and simulate the indoor lighting environment of the area to which the order belongs by using a preset fill light source in a closed environment; The screen images of the display are switched to black, white, red, green and blue in sequence, and a high-definition camera is used to capture the five-color images of the display screen area to generate a night use atlas; The daytime usage atlas and the nighttime usage atlas are packaged to generate a five-color picture set of the display to be tested.

7. The display screen defect detection method based on machine vision according to claim 1, characterized in that: The method of performing defect detection on the set of five-color images to be detected by using a preset visual detection model based on the revised defect detection standard and outputting the defect detection result specifically includes the following steps: Input the corrected defect detection standard into the preset visual inspection model to set the inspection standard; Inputting the five-color images in the five-color image set to be detected into a preset visual detection model, wherein the visual detection model is a convolutional neural network model obtained by deep learning training of standard five-color image sample data of the display; The visual inspection model roughly searches for defects in the five-color image to obtain the ROI area where defects may exist in the five-color image; The visual inspection model performs defect detection on the ROI area of ​​the four-color image according to the defect detection standard and outputs the defect detection results.

8. The display screen defect detection method based on machine vision according to claim 1, characterized in that: The step of preprocessing the five-color picture set to obtain the five-color picture set to be detected specifically comprises the following steps: Filter each color image in the five-color image set to remove texture background and noise points in each color image; Performing sharpening processing on each color image in the five-color image set after filtering, so as to enhance the sharpness and clarity of each color image in the five-color image set; Crop each color picture in the five-color picture set after the sharpening process, remove the environmental background and the display frame of the display to be detected in each color picture, and only keep the picture of the screen area of ​​the display to be detected; The cropped five-color picture set is normalized to obtain a five-color picture set to be detected.

9. A computer-readable storage medium, characterized in that: A computer program is stored which can be loaded by a processor and executes a display screen defect detection method based on machine vision as described in any one of claims 1 to 8.

Images