Method, device and equipment for detecting defects of liquid crystal display screen and storage medium

Abstract

The invention discloses a method, device and equipment for detecting defects of a liquid crystal display screen and a storage medium, relates to the technical field of display panel manufacturing, and aims to reduce the reinspection amount through automatic preliminary screening and perform quantitative analysis based on a defect reinspection priority index Dpr and a reinspection time consumption evaluation value Tce. And then task segmentation of machine quantification and manual judgment is performed on a specific high-time-consumption defect, and deep integration of man-machine advantages is realized. More importantly, the system calculates the real-time bearing capacity ICC by monitoring the seat load in real time, and ensures that the task is accurately matched to the detector in the best state. Through double perception and intelligent matching of defects and personnel, the problems of low efficiency, overload and false detection of traditional detection are fundamentally solved, and the comprehensive performance and robustness of the system are remarkably improved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of display panel manufacturing, in particular to a method and device for detecting defects of a liquid crystal display, equipment and a storage medium. BACKGROUND

[0002] Liquid crystal display screens, as a display technology widely used in current electronic products, have been rapidly developed and popularized due to their high resolution, low power consumption and thinness. However, in the production process of liquid crystal display screens, defect detection as a key link to ensure product quality still faces many technical challenges. The existing defect detection methods for liquid crystal display screens mainly rely on manual visual inspection or automatic detection technology based on traditional image processing algorithms.

[0003] Although the traditional manual detection method has certain flexibility, it is time-consuming and easily affected by human fatigue and subjective judgment, resulting in low detection efficiency and difficulty in ensuring accuracy, which is difficult to meet the needs of large-scale production. This is a common technical drawback in the field of liquid crystal display screen defect detection. To improve detection efficiency, automated detection technology has gradually become a research hotspot. However, existing automatic detection methods mostly rely on fixed thresholds and simple feature extraction, making it difficult to effectively identify diversified and tiny defects, especially under complex background and light change conditions, which is prone to false positives and false negatives, and is an uncommon but major technical drawback that affects the detection results.

[0004] More importantly, the common problem of low detection efficiency directly leads to excessive reliance on and frequent switching of automatic detection technology, while the false positives and false negatives of automatic detection further exacerbate the accuracy decline of overall defect recognition, forming a chain reaction between the two, which seriously restricts the overall performance improvement of liquid crystal display screen defect detection technology and its application in industrial production. SUMMARY

[0005] To overcome the shortcomings of the prior art, the present application provides a method and device for detecting defects of a liquid crystal display, equipment and a storage medium, which solves the technical drawbacks mentioned in the background art.

[0006] To achieve the above purpose, the present application is implemented by the following technical scheme: a method for detecting defects of a liquid crystal display, comprising the following steps: S1, a template image of a standard defect-free liquid crystal display is obtained in advance, a plurality of manual detection positions are locked, and the display screen image and the manual detection positions are numbered; S2, the image of the liquid crystal display to be detected is collected in real time and compared with the template image to generate a preliminary defect set, all suspected defect regions are determined from the image according to the preliminary defect set, and the suspected defect regions are located and preliminarily judged for their automated detection confidence. S3, according to the acquired several suspected defect areas, analyze the image feature information, construct the defect feature set, and generate the defect review priority index Dpr of the suspected defect with serial number i according to the defect feature set, based on the result of the defect review priority index Dpr, substitute the suspected defect with serial number i into the dynamic allocation model, combine its influence data on the production rhythm, predict the review time consumption evaluation value Tce, and judge whether the initially screened suspected defect needs to enter the artificial review process immediately; S4, while monitoring and recording the related working state data information of the artificial detection seat, determine whether the suspected defect with serial number i can be successfully matched to the artificial detection seat with the best state; when the suspected defect with serial number i fails to be successfully matched to the initially selected artificial detection seat, evaluate the next order artificial detection seat again, and obtain the secondary allocation waiting time Sad for comparison, and finally determine whether the suspected defect with serial number i needs to enter the review waiting queue.

[0007] Further, the image of the liquid crystal display to be detected is analyzed in real time, and a preliminary defect set is generated, wherein the preliminary defect set includes the automatic detection confidence ADC of all suspected defect areas, the defect size, and the confidence safety threshold CST set by the system; At the same time, the size relationship between the automatic detection confidence ADC of several suspected defect areas and the confidence safety threshold CST is judged to determine the suspected defects that need to be further analyzed from all image abnormalities, specifically: When the automatic detection confidence ADC≤the confidence safety threshold CST, the corresponding area is determined as a suspected defect at this time, and is marked; When the automatic detection confidence ADC>the confidence safety threshold CST, the corresponding area is directly determined as a real defect by the automatic system at this time.

[0008] Further, according to the several suspected defects obtained in S2, the image feature information is monitored and the defect feature set is constructed; Wherein, the defect feature set includes the contrast C and the shape complexity S of several suspected defects; the defect review priority index Dpr of the suspected defect with serial number i is generated according to the defect feature set, so as to quantify the recognition difficulty of the suspected defect with serial number i for the automatic algorithm; The defect review priority index Dpr with serial number i is obtained by the following formula: ; Wherein, Cᵢ represents the contrast of the suspected defect with serial number i, Sᵢ represents the shape complexity of the suspected defect with serial number i, α and β are the weight coefficients of the corresponding contrast reciprocal and shape complexity respectively, and δ represents the first correction constant.

[0009] Further, in combination with the result of the defect review priority index Dpr of the defect with serial number i, the review order of the suspected defect with serial number i is determined, and the specific content is as follows: When the defect review priority index Dpr of the defect with serial number i is higher than the preset high priority threshold, the priority allocation mode is started, and at this time the suspected defect with current serial number i will be prioritized over other medium and low priority defects for manual seat matching; When the defect review priority index Dpr of the defect with serial number i is in the medium priority range, the normal allocation mode is started, and the matching is performed in the order of entering the queue; When the defect review priority index Dpr of the defect with serial number i is lower than the preset low priority threshold, the low priority mode is started, and the allocation is only performed when the manual detection seat is completely idle.

[0010] Further, the related time consumption information of the suspected defect sent to manual review is evaluated, and the related time consumption information includes the average manual processing time AHT of the defect type and the cycle time CT of the current production line; According to the related time consumption information, the average manual processing time AHT of the corresponding defect and the cycle time CT of the current production line are associated, and the corresponding review time consumption evaluation value Tce is calculated and obtained, and the review time consumption evaluation value Tce is obtained by the following formula: ; Wherein, AHTᵢ represents the average manual processing time of the type corresponding to the i-th suspected defect, and CT represents the cycle time of the current production line.

[0011] Further, according to the review time consumption evaluation value Tce and in combination with the defect review priority index Dpr, it is judged again whether the suspected defect needs to be processed by manual and automatic system task segmentation, and the specific content is as follows: If the review time consumption evaluation value Tce is greater than the preset time influence threshold, and its defect review priority index Dpr is in the medium priority range, it means that the defect review time is too long and may affect the line efficiency, and at this time the system will decompose the review task of the defect, and the automatic system will first complete the deep extraction and quantization of the feature parameters, and then the quantization result will be handed over to the manual for the final quick ruling; If the review time consumption evaluation value Tce is less than or equal to the preset time influence threshold, it means that the defect review time is within the acceptable range, and the complete defect image can be directly allocated to the manual for comprehensive review.

[0012] Further, according to the initially selected artificial detection seat, relevant working state data information of the seat is monitored and acquired, the relevant working state data information including total load index TLI of the seat, assigned task number ATN, idle state IS, and defect number WQN in the waiting queue; according to the relevant working state data information, instant carrying capacity ICC of the artificial detection seat is calculated and acquired, the instant carrying capacity ICC being acquired through the following formula: ; In the formula, IS is a Boolean value, 1 for idle and 0 for busy; TLI is the theoretical maximum task load of the seat; ATN is the assigned task number; WQN is the number of tasks waiting for the seat; whether the suspected defect with serial number i can be successfully matched to the artificial detection seat is further determined according to the re-inspection order of the suspected defect with serial number i and in combination with the instant carrying capacity ICC; when the suspected defect with serial number i can be successfully matched, the system will at this time open an instruction to send a re-inspection task to the artificial detection seat.

[0013] A device for detecting defects of a liquid crystal display screen, comprising: A resource configuration labeling module is configured to pre-acquire a template image of a standard defect-free liquid crystal display screen, lock a plurality of artificial detection seats, and label the display screen image and the artificial detection seats with serial numbers; An image comparison preliminary screening module is configured to acquire images of a liquid crystal display screen to be detected in real time, compare and analyze the images with the template image, generate a preliminary defect set, determine all suspected defect regions from the images according to the preliminary defect set, and locate the suspected defect regions to preliminarily determine the automatic detection confidence thereof; A defect priority determination module is configured to analyze image feature information of a plurality of acquired suspected defect regions, construct a defect feature set, generate a defect re-inspection priority index Dpr of the suspected defect with serial number i according to the defect feature set, based on a result of the defect re-inspection priority index Dpr, substitute the suspected defect with serial number i into a dynamic allocation model, combine its influence data on the production rhythm, predict a re-inspection time consumption evaluation value Tce, and determine whether the initially screened suspected defect needs to immediately enter an artificial re-inspection process; A dynamic task allocation module is configured to simultaneously monitor and record relevant working state data information of artificial detection seats, and then determine whether the suspected defect with serial number i can be successfully matched to an artificial detection seat with the best state; when the suspected defect with serial number i fails to be successfully matched to the initially selected artificial detection seat, a next-order artificial detection seat is re-evaluated, and a secondary allocation waiting time Sad is acquired to be compared, and finally it is determined whether the suspected defect with serial number i needs to enter a re-inspection waiting queue.

[0014] An apparatus for detecting defects of a liquid crystal display screen, comprising a memory and a processor, the memory storing a computer program, the processor being configured to execute the computer program to perform a method for detecting defects of a liquid crystal display screen, the computer program being configured to perform an apparatus for detecting defects of a liquid crystal display screen when executed.

[0015] A storage medium for detecting defects of a liquid crystal display screen, storing a computer program, comprising the following steps, the computer program storing computer instructions, the instructions being executed by a processor to implement a method for detecting defects of a liquid crystal display screen.

[0016] The present application provides a method, device, apparatus and storage medium for detecting defects of a liquid crystal display screen. (1) The method, device, apparatus and storage medium for detecting defects of a liquid crystal display screen effectively solve the defects in the prior art by introducing a set of human-computer collaborative dynamic intelligent detection process. First, the method uses an automatic system to preliminarily screen the collected images, uses automatic detection confidence ADC and a set confidence safety threshold CST to directly confirm high-confidence real defects, greatly reduces the number of defects that need to be manually reviewed, and calculates the defect review priority index Dpr of each suspected defect, which integrates the contrast C and shape complexity S of the defect, realizes the quantitative sorting of the defect review priority, and ensures that the limited manual detection resources can be concentrated to process the most ambiguous and most difficult defects. Further, by evaluating the average manual processing time AHT and the line cycle CT of the defect type to predict the review time consumption evaluation value Tce, the system can predict the potential impact of the review task on production efficiency, thereby realizing fine management of the detection process, completely changing the situation of long time-consuming, low efficiency and being easily affected by subjective fatigue in traditional manual detection, and greatly improving the detection efficiency and objective accuracy.

[0017] (2) The method, device, equipment and storage medium for detecting defects of a liquid crystal display screen, in view of the false detection and missed detection problems caused by the dependence of traditional automatic detection technology on fixed thresholds, and the chain reaction of double decline in efficiency and accuracy caused thereby, propose an innovative dynamic allocation and task segmentation mechanism; this method not only simply hands over difficult defects to manual, but also performs task segmentation processing on specific defects according to the combination of recheck time consumption evaluation value Tce and defect recheck priority index Dpr, i.e. machine first quantizes deeply, and then manual quickly adjudicates, realizing deep integration of man-machine advantages; more importantly, when performing task allocation, the method will monitor the total load index TLI, the number of allocated tasks ATN, the idle state IS and the number of defects in the waiting queue WQN of the manual detection seat in real time, and calculate the instantaneous carrying capacity ICC of each seat accordingly, ensuring that each recheck task can be allocated to the best state and lightest load of the inspector, this double perception and intelligent matching of defect state and inspector state fundamentally breaks the vicious cycle between the rigidity of automatic technology and the overload of manual detection, and significantly improves the robustness and comprehensive performance of the entire detection system in complex production environment. BRIEF DESCRIPTION OF DRAWINGS

[0018] Figure 1 A step flowchart of the method for detecting defects of a liquid crystal display screen according to the present application; Figure 2 A system structure diagram of the system for detecting defects of a liquid crystal display screen according to the present application. DETAILED DESCRIPTION

[0019] The technical solutions in the embodiments of the present application will be described clearly and completely below with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, rather than all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative labor fall within the scope of protection of the present application.

[0020] Embodiment 1 Please refer to Figure 1 The present application provides a method for detecting defects of a liquid crystal display screen, comprising the following steps: S1, pre-acquire a template image of a standard defect-free liquid crystal display screen, lock a plurality of manual detection seats, and mark the display screen image and the manual detection seats with serial numbers; S2, real-time acquisition of the image of the liquid crystal display screen to be detected, comparison and analysis with the template image to generate a preliminary defect set, determination of all suspected defect areas from the image according to the preliminary defect set, positioning to the suspected defect area, and preliminary judgment of the automatic detection confidence; S3, according to the acquired several suspected defect areas, analyze the image feature information, construct a defect feature set, and generate a defect review priority index Dpr of the suspected defect with the serial number i according to the defect feature set, based on the result of the defect review priority index Dpr, the suspected defect with the serial number i is substituted into the dynamic allocation model, combined with the influence data of the production rhythm, the review time consumption evaluation value Tce is predicted, and it is judged whether the initially screened suspected defect needs to enter the artificial review process immediately; S4, the related working state data information of the artificial detection seat is monitored and recorded at the same time, and it is determined whether the suspected defect with the serial number i can be successfully matched to the artificial detection seat with the best state, when the suspected defect with the serial number i cannot be successfully matched to the initially selected artificial detection seat, the next order artificial detection seat is evaluated again, and the secondary allocation waiting time Sad is obtained for comparison, and finally it is determined whether the suspected defect with the serial number i needs to enter the review waiting queue.

[0021] Further, the various step modules in the embodiment of the present application work cooperatively and achieve significant beneficial effects respectively, wherein the resource configuration and labeling step S1 provides a standardized and addressable basis for the entire detection process, which is the prerequisite for subsequent automatic comparison and dynamic task allocation; the image comparison and initial screening step S2 realizes rapid and comprehensive automatic discovery and positioning of defects, greatly improving the efficiency of preliminary detection and preventing early missed detection; the defect priority determination step S3 realizes quantitative evaluation of the necessity and processing cost of suspected defect review by constructing the defect review priority index Dpr and predicting the review time consumption evaluation value Tce, ensuring that valuable artificial review resources can be accurately invested in the defects that need most attention; the dynamic task allocation step S4 realizes dynamic load balancing of detection tasks by real-time monitoring of artificial seat state and intelligent scheduling combined with the secondary allocation waiting time Sad, effectively avoiding the problems of efficiency and accuracy decline caused by fatigue or uneven workload of artificial detection personnel, and finally ensuring efficient, stable and reliable operation of the entire detection system.

[0022] The image of the liquid crystal display screen to be detected is analyzed in real time, and a preliminary defect set is generated, wherein the preliminary defect set includes the automatic detection confidence ADC of all suspected defect areas, the defect size and the confidence safety threshold CST set by the system; At the same time, the size relationship between the automatic detection confidence ADC of several suspected defect areas and the confidence safety threshold CST is judged to determine the suspected defects that need further analysis from all image abnormalities, specifically: When the automatic detection confidence ADC is less than or equal to the confidence safety threshold CST, the corresponding area is determined as a suspected defect and is marked; When the automated detection confidence ADC is greater than the confidence safety threshold CST, the corresponding region is directly determined as a real defect by the automated system at this time.

[0023] According to the several suspected defects obtained in S2, the image feature information thereof is monitored and a defect feature set is constructed; The defect feature set includes the contrast C and the shape complexity S of the several suspected defects; and a defect review priority index Dpr of the suspected defect with the serial number i is generated according to the defect feature set, so as to quantify the identification difficulty of the suspected defect with the serial number i for the automated algorithm. The defect review priority index Dpr of the defect with the serial number i is obtained by the following formula: ; Wherein, Cᵢ represents the contrast of the suspected defect with the serial number i, Sᵢ represents the shape complexity of the suspected defect with the serial number i, α and β are the weight coefficients of the corresponding contrast reciprocal and shape complexity respectively, and δ represents the first correction constant.

[0024] In combination with the result of the defect review priority index Dpr of the defect with the serial number i, the review order of the suspected defect with the serial number i is determined, and the specific content is as follows: When the defect review priority index Dpr of the defect with the serial number i is higher than the preset high priority threshold, the priority allocation mode is started, and the suspected defect with the current serial number i will be prioritized over other medium and low priority defects for artificial seat matching; When the defect review priority index Dpr of the defect with the serial number i is in the medium priority range, the normal allocation mode is started, and the matching is performed according to the order of entering the queue; When the defect review priority index Dpr of the defect with the serial number i is lower than the preset low priority threshold, the low priority mode is started, and the allocation is only performed when the artificial detection seat is completely idle.

[0025] Meanwhile, the related time consumption information of the suspected defect sent to the artificial review is evaluated, and the related time consumption information includes the average artificial processing time AHT of the defect type and the cycle time CT of the current production line. According to the related time consumption information, the average artificial processing time AHT of the corresponding defect and the cycle time CT of the current production line are associated, and a corresponding review time consumption evaluation value Tce is calculated and obtained, and the review time consumption evaluation value Tce is obtained by the following formula: ; Wherein, AHTᵢ represents the average artificial processing time of the type corresponding to the i-th suspected defect, and CT represents the cycle time of the current production line.

[0026] The significance of the calculation formula of the re-inspection time consumption evaluation value Tce is to predict the relative impact of the manual re-inspection process of the current suspected defect i on the production cycle. Furthermore, the constructed defect quantification and prioritization module plays a core role in intelligent decision-making, connecting the entire inspection system. Its significance lies in providing a multi-dimensional profile of defects by collecting a series of key lower-level parameters. Specifically, the significance of collecting the automated detection confidence ADC and setting the confidence safety threshold CST is to establish the first automated screening barrier, quickly separating clear defects from suspected defects, thereby greatly reducing the burden of subsequent analysis. The significance of collecting the defect contrast C and shape complexity S is to capture the ambiguity and irregularity of the defect from the perspective of image physical characteristics, providing the original basis for evaluating its recognition difficulty. The significance of collecting the average manual processing time (AHT) and the current production line's cycle time (CT) for each defect type lies in obtaining data on the time cost and efficiency impact of re-inspection tasks from a production management perspective. Based on these parameters, in-depth system evaluation is also crucial. Calculating the defect re-inspection priority index (Dpr) through a formula transforms the elusive "identification difficulty" into a quantifiable and sortable value, providing a scientific and unique basis for subsequent priority queue allocation. Ultimately, this complete set of logic, from multi-dimensional parameter collection to composite indicator evaluation, forms the cornerstone of the graded response mechanism, ensuring that detection resources can be optimally allocated based on the inherent difficulty of the defect and the external impact cost, thereby achieving a dual improvement in efficiency and accuracy.

[0027] Based on the re-inspection time consumption evaluation value Tce and the defect re-inspection priority index Dpr, it is determined again whether the suspected defect requires task division between manual and automated systems. The specific contents are as follows: If the re-inspection time consumption evaluation value Tce is greater than the preset time impact threshold, and the defect re-inspection priority index Dpr is in the medium priority range, it means that the defect re-inspection is taking too long and may affect the production line efficiency. In this case, the system will decompose the re-inspection task of the defect. The automated system will first complete the in-depth extraction and quantification of the characteristic parameters, and then hand over the quantification results to the human for final and rapid judgment. If the re-inspection time consumption evaluation value Tce is less than or equal to the preset time impact threshold, it indicates that the defect re-inspection time consumption is within an acceptable range, and the complete defect image can be directly assigned to manual re-inspection for comprehensive re-inspection.

[0028] According to the initially selected artificial detection seat, relevant working state data information of the artificial detection seat is monitored and acquired, the relevant working state data information including total load index TLI, assigned task number ATN, idle state IS and defect number WQN in the waiting queue of the seat; according to the relevant working state data information, the instant carrying capacity ICC of the artificial detection seat is calculated and acquired, the instant carrying capacity ICC being acquired through the following formula: ; In the formula, IS is a Boolean value, 1 for idle and 0 for busy; TLI is the theoretical maximum task load of the seat; ATN is the assigned task number; WQN is the number of tasks waiting for the seat; whether the suspected defect with serial number i can be successfully matched to the artificial detection seat is further determined according to the re-inspection order of the suspected defect with serial number i and in combination with the instant carrying capacity ICC; when the suspected defect with serial number i can be successfully matched, the system will start the instruction of sending a re-inspection task to the artificial detection seat.

[0029] Further, a secondary deep decision is made based on the previously evaluated re-inspection time consumption evaluation value Tce and the defect re-inspection priority index Dpr; the meaning of this decision is that when a defect is determined to be time-consuming and moderately difficult, the system can actively perform task segmentation, that is, the machine shares the burden of deep analysis and the human completes the final ruling, which is the most efficient optimization combination of human and machine resources, aiming to ensure the smoothness of the production line rhythm; At the same time, the real-time state of the detection employee is accurately perceived by collecting a series of lower-level parameters of the artificial seat, wherein the collection of the total load index TLI sets the theoretical upper limit, the collection of the assigned task number ATN and the defect number WQN in the waiting queue grasps the current and future actual load, and the collection of the idle state IS provides the most direct availability judgment; the deep meaning of substituting these parameters into the formula to calculate the instant carrying capacity ICC is that it integrates multiple dimensional discrete state data into a unified and dynamic quantitative index, so that the system can objectively compare and optimally allocate the carrying capacity of different detection employees; ultimately, this dual deep insight into the characteristics and processing method of the defect and the real-time state of the detection employee and intelligent matching fundamentally solve the blindness of task allocation, realize real load balancing, and ensure that the detection process can still maintain the highest efficiency and reliability when dealing with complex and variable production conditions.

[0030] Embodiment 2 Please refer to Figure 2 A device for detecting defects of a liquid crystal display screen, comprising: A resource configuration marking module is configured to pre-acquire a template image of a standard defect-free liquid crystal display screen, lock a plurality of artificial detection seats, and mark the display screen image and the artificial detection seats with serial numbers. An image comparison preliminary screening module is configured to capture images of liquid crystal display screens to be detected in real time, and compare and analyze the images with the template images to generate a preliminary defect set, determine all suspected defect regions from the images according to the preliminary defect set, and locate the suspected defect regions to preliminarily determine the automated detection confidence thereof; A defect priority determination module is configured to analyze image feature information of the obtained suspected defect regions, construct a defect feature set, and generate a defect review priority index Dpr of the suspected defect with the serial number i according to the defect feature set, and based on the result of the defect review priority index Dpr, substitute the suspected defect with the serial number i into a dynamic allocation model, combine the influence data of the suspected defect on the production rhythm, predict a review time consumption evaluation value Tce, and determine whether the suspected defect preliminarily screened needs to immediately enter a manual review process. A dynamic task allocation module is configured to simultaneously monitor and record relevant working state data information of manual detection seats, and then determine whether the suspected defect with the serial number i can be successfully matched to a manual detection seat with the best state; when the suspected defect with the serial number i fails to be successfully matched to the preliminarily selected manual detection seat, the next manual detection seat is re-evaluated, and a secondary allocation waiting time Sad is obtained for comparison, and finally it is determined whether the suspected defect with the serial number i needs to enter a review waiting queue.

[0031] An apparatus for detecting defects of liquid crystal display screens, comprising a memory and a processor, the memory stores a computer program, and the processor is configured to run the computer program to execute a method for detecting defects of liquid crystal display screens, and the computer program is configured to execute an apparatus for detecting defects of liquid crystal display screens when running.

[0032] A storage medium for detecting defects of liquid crystal display screens, which stores a computer program, comprising the following steps, the computer program stores computer instructions, and the instructions are executed by a processor to realize a method for detecting defects of liquid crystal display screens.

[0033] Although embodiments of the present application have been shown and described, it is to be understood that various modifications, substitutions, replacements and changes can be made to these embodiments without departing from the principles and spirit of the present application, and the scope of the present application is defined by the appended claims and their equivalents.

Claims

1. A method for detecting defects in a liquid crystal display screen, characterized in that: The following steps are involved: S1. Obtain a template image of a standard, defect-free LCD screen in advance, lock in several manual inspection seats, and mark the screen image and the manual inspection seats with serial numbers; S2. Real-time acquisition of an image of the LCD screen to be inspected, and comparison analysis with the template image to generate a preliminary defect set, determine all suspected defect areas from the image based on the preliminary defect set, locate the suspected defect areas, and preliminarily determine the confidence level of automated inspection; S3. Analyze the image feature information of the acquired suspected defect areas to construct a defect feature set. Generate a defect re-inspection priority index Dpr for the suspected defect numbered i based on the defect feature set. Based on the defect re-inspection priority index Dpr, substitute the suspected defect numbered i into the dynamic allocation model. Combined with the data on its impact on the production cycle, predict the re-inspection time consumption assessment value Tce to determine whether the initially screened suspected defects need to enter the manual re-inspection process immediately. S4. Simultaneously, the relevant working status data information of the manual inspection seat is monitored and recorded, and then it is determined whether the suspected defect with serial number i can be successfully matched to the manual inspection seat with the best status; when the suspected defect with serial number i fails to be successfully matched to the initially selected manual inspection seat, the next-ranked manual inspection seat is evaluated again, and the secondary allocation waiting time Sad is obtained for comparison, and finally it is determined whether the suspected defect with serial number i needs to enter the re-inspection waiting queue.

2. The method for detecting defects in a liquid crystal display screen according to claim 1, wherein: Performing real-time analysis on the image of the LCD to be inspected and generating a preliminary defect set, wherein the preliminary defect set includes the automated detection confidence ADC, defect size, and system-set confidence safety threshold CST corresponding to all suspected defect areas; At the same time, the relationship between the automated detection confidence ADC and the confidence safety threshold CST of several suspected defect areas is determined to determine the suspected defects that require further analysis from all image anomalies. Specifically: When the automated detection confidence ADC ≤ the confidence safety threshold CST, the corresponding area is identified as a suspected defect and marked; When the automated detection confidence ADC is greater than the confidence safety threshold CST, the corresponding area is directly determined by the automated system as a real defect.

3. The method for detecting defects in a liquid crystal display screen according to claim 1, wherein: Based on the suspected defects obtained in S2, monitor their image feature information and construct a defect feature set; The defect feature set includes the contrast C and shape complexity S of several suspected defects; a defect re-inspection priority index Dpr of the suspected defect with sequence number i is generated based on the defect feature set to quantify the difficulty of identifying the suspected defect with sequence number i for the automated algorithm; The defect re-inspection priority index Dpr of the sequence number i is obtained by the following formula: ； Where Cᵢ represents the contrast of the suspected defect with sequence number i, Sᵢ represents the shape complexity of the suspected defect with sequence number i, α and β are the corresponding weight coefficients of the inverse of contrast and shape complexity, respectively, and δ represents the first correction constant.

4. The method for detecting defects in a liquid crystal display screen according to claim 1, wherein: Combined with the result of the defect re-inspection priority index Dpr of sequence number i, the re-inspection order of the suspected defect of sequence number i is determined. The specific contents are as follows; When the defect re-inspection priority index Dpr of the defect with sequence number i is higher than the preset high priority threshold, the priority allocation mode is enabled. At this time, the suspected defect with sequence number i will be prioritized over other medium and low priority defects for manual seat matching. When the defect re-inspection priority index Dpr of the defect numbered i is in the medium priority range, the normal allocation mode is enabled and matching is performed in the order in which the defects enter the queue. When the defect re-inspection priority index Dpr of the defect with serial number i is lower than the preset low priority threshold, the low priority mode is turned on and the manual inspection seat is only allocated when it is completely idle.

5. The method for detecting defects in a liquid crystal display screen according to claim 1, wherein: At the same time, the relevant time consumption information of suspected defects sent for manual re-inspection is evaluated. The relevant time consumption information includes the average manual processing time (AHT) of the defect type and the tact cycle (CT) of the current production line; Based on the relevant time consumption information, the average manual processing time AHT of the corresponding defect and the takt cycle CT of the current production line are correlated to calculate the corresponding re-inspection time consumption evaluation value Tce. The re-inspection time consumption evaluation value Tce is obtained by the following formula: ； Where AHTᵢ represents the average manual processing time of the corresponding type of the i-th suspected defect, and CT represents the takt time of the current production line.

6. The method for detecting defects in a liquid crystal display screen according to claim 1, wherein: Based on the re-inspection time consumption evaluation value Tce and the defect re-inspection priority index Dpr, it is determined again whether the suspected defect requires task division between manual and automated systems. The specific contents are as follows: If the re-inspection time consumption evaluation value Tce is greater than the preset time impact threshold, and the defect re-inspection priority index Dpr is in the medium priority range, it means that the defect re-inspection is taking too long and may affect the production line efficiency. In this case, the system will decompose the re-inspection task of the defect. The automated system will first complete the in-depth extraction and quantification of the characteristic parameters, and then hand over the quantification results to the human for final and rapid judgment. If the re-inspection time consumption evaluation value Tce is less than or equal to the preset time impact threshold, it indicates that the defect re-inspection time consumption is within an acceptable range, and the complete defect image can be directly assigned to manual re-inspection for comprehensive re-inspection.

7. The method for detecting defects in a liquid crystal display screen according to claim 1, wherein: Based on the preliminarily selected manual inspection seat, relevant work status data information is monitored and obtained. The relevant work status data information includes the total load index TLI, the number of assigned tasks ATN, the idle state IS, and the number of defects in the waiting queue WQN of the seat. Based on the relevant work status data information, the instantaneous carrying capacity ICC of the manual inspection seat is calculated and obtained. The instantaneous carrying capacity ICC is obtained by the following formula: ； Where IS is a Boolean value, with idle being 1 and busy being 0; TLI is the theoretical maximum task load of the seat; ATN is the number of assigned tasks; WQN is the number of tasks waiting for the seat; the re-inspection order of the suspected defect with sequence number i is combined with the instantaneous carrying capacity ICC to further determine whether the suspected defect with sequence number i can be successfully matched to the manual inspection seat; when the suspected defect with sequence number i can be successfully matched, the system will start sending the re-inspection task to the manual inspection seat.

8. A device for detecting defects in a liquid crystal display screen, comprising the device for detecting defects in a liquid crystal display screen according to any one of claims 1 to 7, characterized in that: Resource configuration and labeling module, used to pre-acquire template images of standard defect-free LCD screens, lock several manual inspection seats, and label the screen images and manual inspection seats with serial numbers; An image comparison and preliminary screening module is used to collect images of the LCD screen to be inspected in real time and compare and analyze them with the template image to generate a preliminary defect set. Based on the preliminary defect set, all suspected defect areas are determined from the image, and the suspected defect areas are located to preliminarily determine their automated inspection confidence. The defect priority determination module is used to analyze the image feature information of several acquired suspected defect areas, construct a defect feature set, and generate a defect re-inspection priority index Dpr for the suspected defect with a serial number i based on the defect feature set. Based on the result of the defect re-inspection priority index Dpr, the suspected defect with a serial number i is substituted into the dynamic allocation model, and combined with the data on its impact on the production cycle, the re-inspection time consumption evaluation value Tce is predicted to determine whether the suspected defects initially screened need to enter the manual re-inspection process immediately; The dynamic task allocation module is used to simultaneously monitor and record the relevant work status data information of the manual inspection seat, and then determine whether the suspected defect with serial number i can be successfully matched to the manual inspection seat with the best status; when the suspected defect with serial number i fails to successfully match the initially selected manual inspection seat, the next-ranked manual inspection seat is evaluated again, and the secondary allocation waiting time Sad is obtained for comparison, and finally it is determined whether the suspected defect with serial number i needs to enter the re-inspection waiting queue.

9. A device for detecting defects in a liquid crystal display screen, comprising a memory and a processor, characterized in that: The memory stores a computer program, the processor is configured to run the computer program to execute a method for detecting defects in a liquid crystal display screen according to any one of claims 1 to 7, and the computer program is configured to execute a device for detecting defects in a liquid crystal display screen according to claim 8 when running.

10. A storage medium for detecting defects in a liquid crystal display screen, having a computer program stored thereon, characterized in that: The method comprises the following steps: the computer program stores computer instructions, and when the instructions are executed by a processor, a method for detecting defects of a liquid crystal display screen according to any one of claims 1 to 7 is implemented.

Images