Microcrack defect detection method, system and storage medium for inner arc edge of curved screen

Abstract

The present invention discloses a method, system, and storage medium for detecting microcrack defects on the inner arc edge of a curved screen. The method comprises: obtaining multiple inner arc edge images corresponding to each inner arc edge of the curved screen to be inspected; wherein the multiple inner arc edge images corresponding to any inner arc edge include a complete image of the inner arc edge; inputting each inner arc edge image into a target defect detection model for detecting microcrack areas, obtaining a microcrack defect detection result for each inner arc edge image; and determining the inner arc edge defect area of ​​the curved screen to be inspected based on all microcrack defect detection results. The present invention improves the accuracy of microcrack defect detection on the inner arc edge of a curved screen while also improving detection efficiency.

Description

Technical Field

[0001] The present invention relates to the technical field of pipeline control, and in particular to a method, system and storage medium for detecting micro-crack defects on the inner arc edge of a curved screen. Background Art

[0002] A crucial step in producing curved screens is laminating the curved glass to the display and other components. During this lamination process, the edges are prone to peeling, damage, or compression from other components during lamination, resulting in damage to the display. Air gradually intrudes into these damaged areas and continuously oxidizes organic matter on the screen, ultimately creating growing dark spots (GDS). The development of these dark spots has a certain lag, and they may not form until after factory quality inspection, leading to customer complaints and becoming a critical issue that display manufacturers urgently need to address in their quality inspection process.

[0003] Currently, quality inspections for GDS are generally divided into three types: screen hole area inspection, screen inner arc edge area inspection, and screen outer arc edge area inspection. The defects in the screen inner arc edge area are mainly cracks, and the crack width is usually 1 to 3 microns. In addition, since the display substrate exists on the surface of the inner arc edge after the bonding process, general optical imaging equipment will include interference such as scratches and dirt on the substrate in the image, making it difficult to distinguish from cracks. However, display screen manufacturers currently do not conduct automated inspections for this type of defect and can only conduct manual spot checks.

[0004] Therefore, it is urgent to provide a technical solution to solve the above problems. Summary of the Invention

[0005] In order to solve the above technical problems, the present invention provides a method, system and storage medium for detecting micro-crack defects on the inner arc edge of a curved screen.

[0006] The technical solution of the method for detecting micro-crack defects on the inner arc edge of a curved screen of the present invention is as follows:

[0007] Acquire multiple inner arc edge images corresponding to each inner arc edge of the curved screen to be inspected; wherein the multiple inner arc edge images corresponding to any inner arc edge include a complete image of the inner arc edge;

[0008] Inputting each inner arc edge image into a target defect detection model for detecting microcrack areas, respectively, to obtain a microcrack defect detection result for each inner arc edge image;

[0009] Based on all microcrack defect detection results, the inner arc edge defect area of ​​the curved screen to be detected is determined.

[0010] The beneficial effects of the method for detecting micro-crack defects on the inner arc edge of a curved screen of the present invention are as follows:

[0011] The method of the present invention not only improves the accuracy of micro-crack defect detection on the inner arc edge of the curved screen, but also improves the detection efficiency.

[0012] Based on the above solution, the method for detecting micro-crack defects on the inner arc edge of a curved screen of the present invention can be further improved as follows.

[0013] Furthermore, the step of obtaining a plurality of inner arc edge images corresponding to each inner arc edge of the curved screen to be inspected includes:

[0014] A plurality of inner arc edge images corresponding to each inner arc edge are acquired by using a microscope camera with a half-pupil focusing technique.

[0015] The beneficial effect of adopting the above-mentioned further technical solution is that the microscope camera uses half-pupil focusing technology to quickly focus, focusing the lens on the TFT circuit layer inside the curved screen, blurring surface interference to distinguish interference from microcracks, thereby providing more accurate images for subsequent defect detection.

[0016] Furthermore, the target defect detection model includes: a target image segmentation module, a target image classification module, and a feature extraction module that are sequentially connected and trained; the step of inputting each inner arc edge image into the target defect detection model for detecting the microcrack area to obtain the microcrack defect detection result of each inner arc edge image includes:

[0017] Input any inner arc edge image into the target image segmentation module, obtain the image segmentation features of the inner arc edge image and input them into the target image classification module, obtain the image classification features of the inner arc edge image and input them into the feature extraction module, and obtain the microcrack defect detection result containing the microcrack defect features of the inner arc edge image, until the microcrack defect detection result of each inner arc edge image is obtained.

[0018] Furthermore, the target image segmentation module includes: a first semantic feature extraction layer, a second semantic feature extraction layer, a third semantic feature extraction layer, a fourth semantic feature extraction layer, a fifth semantic feature extraction layer, a sixth semantic feature extraction layer, a fusion layer, and a seghead layer, and a first detail feature extraction layer, a second detail feature extraction layer, and a third detail feature extraction layer, which are arranged in sequence; the step of inputting any inner arc edge image into the target image segmentation module to obtain the image segmentation features of the inner arc edge image includes:

[0019] Inputting any inner arc edge image into the first semantic feature extraction layer and sequentially performing semantic feature extraction on the second semantic feature extraction layer and the third semantic feature extraction layer to obtain a first intermediate feature of the inner arc edge image;

[0020] Inputting the first intermediate feature of any inner arc edge image into the fourth semantic feature extraction layer and the first detail feature extraction layer respectively, so as to sequentially pass the first intermediate feature input into the fourth semantic feature extraction layer through the fifth semantic feature extraction layer and the sixth semantic feature extraction layer for feature extraction to obtain a second intermediate feature, and sequentially pass the first intermediate feature input into the first detail feature extraction layer through the second detail feature extraction layer and the third detail feature extraction layer to obtain a third intermediate feature; wherein, feature fusion is performed respectively between the fourth semantic feature extraction layer and the first detail feature extraction layer, between the fifth semantic feature extraction layer and the second detail feature extraction layer, and between the sixth semantic feature extraction layer and the third detail feature extraction layer;

[0021] The second intermediate feature and the third intermediate feature of any inner arc edge image are input into the fusion layer for feature fusion to obtain the target fusion feature and input into the seghead layer for image segmentation processing to obtain the image segmentation feature of the inner arc edge image.

[0022] Furthermore, the target image classification module includes: sequentially connecting a plurality of re-parameterized residual structures; inputting the image segmentation features of any inner arc edge image into the target image classification module to obtain the image classification features of the inner arc edge image, including:

[0023] The image segmentation features of any inner arc edge image are input into the first re-parameterized residual structure and image classification processing is performed in turn through each re-parameterized residual structure to obtain the image classification features of the inner arc edge image.

[0024] Furthermore, the step of determining the inner arc edge defect area of ​​the curved screen to be inspected based on all microcrack defect detection results includes:

[0025] The inner arc edge defect area of ​​the curved screen to be inspected is determined according to the location of the microcrack defect feature corresponding to each microcrack defect detection result.

[0026] Furthermore, it also includes:

[0027] When the area of ​​the inner arc edge defect region is larger than a preset defect area, the curved screen to be inspected is determined to be unqualified and an alarm message is output.

[0028] The technical solution of the micro-crack defect detection system for the inner arc edge of the curved screen of the present invention is as follows:

[0029] Includes: acquisition module, detection module and operation module;

[0030] The acquisition module is used to: acquire multiple inner arc edge images corresponding to each inner arc edge of the curved screen to be detected; wherein the multiple inner arc edge images corresponding to any inner arc edge include a complete image of the inner arc edge;

[0031] The detection module is used to: input each inner arc edge image into a target defect detection model for detecting microcrack areas, and obtain a microcrack defect detection result of each inner arc edge image;

[0032] The operation module is used to determine the inner arc edge defect area of ​​the curved screen to be detected based on all microcrack defect detection results.

[0033] The beneficial effects of the micro-crack defect detection system for the inner arc edge of a curved screen of the present invention are as follows:

[0034] The system of the present invention not only improves the accuracy of micro-crack defect detection on the inner arc edge of the curved screen, but also improves the detection efficiency.

[0035] Based on the above solution, the micro-crack defect detection system for the inner arc edge of the curved screen of the present invention can also be improved as follows.

[0036] Furthermore, the acquisition module is specifically used to:

[0037] A plurality of inner arc edge images corresponding to each inner arc edge are acquired by using a microscope camera with a half-pupil focusing technique.

[0038] The beneficial effect of adopting the above-mentioned further technical solution is that the microscope camera uses half-pupil focusing technology to quickly focus, focusing the lens on the TFT circuit layer inside the curved screen, blurring surface interference to distinguish interference from microcracks, thereby providing more accurate images for subsequent defect detection.

[0039] A technical solution of a storage medium of the present invention is as follows:

[0040] The storage medium stores instructions, and when a computer reads the instructions, the computer is caused to execute the steps of the method for detecting micro-crack defects in the inner arc edge of a curved screen of the present invention. BRIEF DESCRIPTION OF THE DRAWINGS

[0041] Figure 1 A schematic flow chart showing a first embodiment of a method for detecting micro-crack defects on an inner arc edge of a curved screen according to the present invention is shown;

[0042] Figure 2 A schematic diagram showing the principle of a microscope camera in a first embodiment of a method for detecting micro-crack defects on an inner arc edge of a curved screen according to the present invention;

[0043] Figure 3A schematic diagram showing the principle of the half-pupil focusing technology in the first embodiment of the method for detecting micro-crack defects on the inner arc edge of a curved screen of the present invention;

[0044] Figure 4 A schematic diagram showing the structure of a target image segmentation module in a first embodiment of a method for detecting micro-crack defects on an inner arc edge of a curved screen according to the present invention is shown;

[0045] Figure 5 A schematic diagram showing the structure of the original BiseNet-V2 model in the first embodiment of the method for detecting micro-crack defects on the inner arc edge of a curved screen according to the present invention is shown;

[0046] Figure 6 A schematic diagram showing the structure of a target image classification module in a first embodiment of a method for detecting micro-crack defects on an inner arc edge of a curved screen according to the present invention is shown;

[0047] Figure 7 A schematic diagram showing a residual structure in a first embodiment of a method for detecting micro-crack defects on an inner arc edge of a curved screen according to the present invention;

[0048] Figure 8 A schematic diagram showing the Rep structure in the first embodiment of the method for detecting micro-crack defects on the inner arc edge of a curved screen according to the present invention;

[0049] Figure 9 A schematic diagram showing the principle of a detection device in a second embodiment of a method for detecting micro-crack defects on an inner arc edge of a curved screen according to the present invention;

[0050] Figure 10 The figure shows a schematic structural diagram of an embodiment of a system for detecting micro-crack defects on an inner arc edge of a curved screen provided by the present invention. DETAILED DESCRIPTION

[0051] Figure 1 The figure shows a flow chart of an embodiment of the method for detecting micro crack defects on the inner arc edge of a curved screen provided by the present invention. Figure 1 As shown, the following steps are included:

[0052] Step 110: Acquire multiple inner arc edge images corresponding to each inner arc edge of the curved screen to be inspected.

[0053] Among them, ① the curved screen to be tested is: a curved screen used in a smart terminal and requiring testing. Curved screens include: 2.5D curved screens and 3D curved screens. ② The inner arc edge is: the inner arc edge of the curved screen. A curved screen contains two long sides and two short sides. ③ The inner arc edge image is: an image containing a partial length (image) of a corresponding inner arc edge. ④ The multiple inner arc edge images corresponding to any inner arc edge contain a complete image of the inner arc edge, that is, the multiple inner arc edge images corresponding to any inner arc edge are spliced ​​together to obtain a complete image of the inner arc edge.

[0054] It should be noted that the defects in the inner arc edge area of ​​the curved screen are mainly microcracks, and the crack width is usually 1-3 microns.

[0055] Step 120: Input each inner arc edge image into a target defect detection model for detecting microcrack areas, and obtain a microcrack defect detection result for each inner arc edge image.

[0056] ① The target defect detection model is a pre-trained deep learning model for detecting microcrack areas. It includes a target image segmentation module, a target image classification module, and a feature extraction module, all connected and trained in sequence. ② The microcrack defect detection result is determined by determining whether the inner arc edge image contains microcrack defects. If so, the inner arc edge image also includes the defect feature image corresponding to the microcrack defect.

[0057] Step 130: Based on all microcrack defect detection results, determine the inner arc edge defect area of ​​the curved screen to be detected.

[0058] The inner arc edge defect area is the area corresponding to all microcrack defects in the entire curved screen to be inspected.

[0059] Preferably, step 110 includes:

[0060] A plurality of inner arc edge images corresponding to each inner arc edge are acquired by using a microscope camera with a half-pupil focusing technique.

[0061] It should be noted that ① the microscope camera uses a 5-10x microscope lens, so that the single pixel corresponding size accuracy can reach up to 0.5 microns. In this embodiment, a 7.5x microscope lens is used by default, so that the single pixel corresponding size accuracy reaches 0.64 microns, which can well image a crack with a width of 1 micron. Figure 2 As shown, the microscope camera adopts a 3-axis movable structure, in which the YZ axes are translation and the X axis is rotation. The X-axis rotation is to adapt to curved screens with different curvature angles, so that the camera imaging is parallel to the area to be inspected. Due to the movement interference between the lens and the screen, the mass production equipment can adapt to curved screens with a curvature of 0 to 60°; the Y-axis translation is to adapt to curved screens with different screen widths; the Z-axis translation is to adapt to curved screens with different screen thicknesses, and when the alignment fails, the module can be raised to prevent damage to the curved screen caused by touching it. ② Since the surface of the inner arc edge of the curved screen to be inspected is also covered with a layer of substrate, the microscope camera uses half-pupil focusing technology to focus on the inner layer of the inner arc edge of the screen. The principle of half-pupil focusing technology is as follows Figure 3 As shown, specifically: the autofocus emits a semicircular oblique laser beam. When it is exactly at the focus position, the laser will become a point shape, otherwise it will still be not semicircular and needs to be focused.

[0062] Preferably, step 120 includes:

[0063] Input any inner arc edge image into the target image segmentation module, obtain the image segmentation features of the inner arc edge image and input them into the target image classification module, obtain the image classification features of the inner arc edge image and input them into the feature extraction module, and obtain the microcrack defect detection result containing the microcrack defect features of the inner arc edge image, until the microcrack defect detection result of each inner arc edge image is obtained.

[0064] Preferably, if Figure 4 As shown, the target image segmentation module includes: a first semantic feature extraction layer, a second semantic feature extraction layer, a third semantic feature extraction layer, a fourth semantic feature extraction layer, a fifth semantic feature extraction layer, a sixth semantic feature extraction layer, a fusion layer and a seghead layer, as well as a first detail feature extraction layer, a second detail feature extraction layer and a third detail feature extraction layer, which are arranged in sequence.

[0065] The step of inputting any inner arc edge image into the target image segmentation module to obtain image segmentation features of the inner arc edge image includes:

[0066] Any inner arc edge image is input into the first semantic feature extraction layer and sequentially passed through the second semantic feature extraction layer and the third semantic feature extraction layer for semantic feature extraction to obtain a first intermediate feature of the inner arc edge image.

[0067] The first intermediate feature of any inner arc edge image is respectively input into the fourth semantic feature extraction layer and the first detail feature extraction layer, so that the first intermediate feature input into the fourth semantic feature extraction layer is sequentially passed through the fifth semantic feature extraction layer and the sixth semantic feature extraction layer for feature extraction to obtain a second intermediate feature, and the first intermediate feature input into the first detail feature extraction layer is sequentially passed through the second detail feature extraction layer and the third detail feature extraction layer to obtain a third intermediate feature.

[0068] The second intermediate feature and the third intermediate feature of any inner arc edge image are input into the fusion layer for feature fusion to obtain the target fusion feature and input into the seghead layer for image segmentation processing to obtain the image segmentation feature of the inner arc edge image.

[0069] ① Feature fusion is performed between the fourth semantic feature extraction layer and the first detail feature extraction layer, the fifth semantic feature extraction layer and the second detail feature extraction layer, and the sixth semantic feature extraction layer and the third detail feature extraction layer. ② The target image segmentation module is an improvement based on the lightweight segmentation model BiseNet-V2. ③ The first semantic feature extraction layer includes 32 convolution kernels and extracts 1 / 2 resolution features. The second semantic feature extraction layer includes 32 convolution kernels and extracts 1 / 4 resolution features. The third semantic feature extraction layer includes 64 convolution kernels and extracts 1 / 8 resolution features. The fourth semantic feature extraction layer includes 128 convolution kernels and extracts 1 / 16 resolution features. The fifth semantic feature extraction layer includes 256 convolution kernels and extracts 1 / 32 resolution features. The sixth semantic feature extraction layer includes 512 convolution kernels and extracts 1 / 64 resolution features. The first detail feature extraction layer includes 128 convolution kernels and extracts 1 / 8 resolution features by using 128 convolution kernels. The second detail feature extraction layer includes 128 convolution kernels and extracts 1 / 8 resolution features by using 128 convolution kernels. The third detail feature extraction layer includes 128 convolution kernels and extracts 1 / 8 resolution features by using 128 convolution kernels.

[0070] It should be noted that Figure 5The structural diagram of the original BiseNet-V2 model is shown. Since the detail feature branch extracts high-resolution features from the image, although it uses a shallow layer, that is, only feature extraction at 1 / 2, 1 / 4, and 1 / 8 resolutions is performed, the branch has a large amount of computation due to the use of wide channels to extract features, accounting for about 1 / 3 of the entire model. Since microcracks are prone to breakage and disappear during imaging, and there is a lot of interference on the substrate surface, high-level semantic features at a resolution of 1 / 32 are no longer sufficient. Therefore, in the improved BiseNet-V2 model (target image segmentation module), high-level semantic features are extracted using a resolution of 1 / 64. At the same time, in order to ensure strict production rhythm, the detail feature branch is optimized to use partial features of the semantic feature branch for extraction. The feature extraction structure of the 1 / 2, 1 / 4, and 1 / 8 resolutions in the semantic feature branch is directly used. Feature extraction at a resolution of 1 / 64 is added at the end of the semantic feature branch. Although the detail feature branch uses some of the features extracted at 1 / 8 resolution in the semantic feature branch, after three subsequent fusions of 1 / 8 resolution and lower resolution, its detail feature extraction surpasses that of BiseNet-V2's detail feature branch. Furthermore, because all operations are performed at lower resolutions, the total computational effort is also less than that of BiseNet-V2's detail feature branch. The final target image segmentation module's total runtime was reduced by 15%, and the F1-score accuracy for microcracks was improved by 10.4%.

[0071] Among them, ① wide channel refers to the use of a larger number of convolution kernels for feature extraction. For example, 64 convolution kernels are used to extract 1 / 4 resolution features, and 128 are used for 1 / 8 resolution. ② shallow layer refers to the level at which features are extracted to 1 / 8 resolution. ③ narrow channel refers to the use of a smaller number of convolution kernels for feature extraction. For example, 32 convolution kernels are used to extract 1 / 4 resolution features, and 64 are used for 1 / 8 resolution. Note: 1 / 2 and 1 / 4 use the same number of convolution kernels, and subsequent low-resolution features such as 1 / 8 resolution are each times the number of convolution kernels before * 2. ④ deeper layer refers to the level at which features are extracted to 1 / 32 resolution.

[0072] Preferably, if Figure 6 As shown, the target image classification module includes: a plurality of re-parameterized residual structures connected in sequence.

[0073] The step of inputting the image segmentation features of any inner arc edge image into the target image classification module to obtain the image classification features of the inner arc edge image comprises:

[0074] The image segmentation features of any inner arc edge image are input into the first re-parameterized residual structure and image classification processing is performed in turn through each re-parameterized residual structure to obtain the image classification features of the inner arc edge image.

[0075] It should be noted that ① Although the target image segmentation module can filter out some interference (such as indentations, hair, and dust) during detection, some interference that is difficult to distinguish (such as tiny scratches) will also be detected as defects by the target image segmentation module. In this case, the target image classification module is required to perform a secondary judgment to ensure that as few over-inspections as possible. ② The target image classification module is obtained by using structural reparameterization of the VGG model. During model training, Figure 7 The residual structure shown is rewritten as Figure 8 The Rep structure shown in the figure makes the model have better accuracy. During model inference, the 1*1 convolution and residual structure branches are reparameterized and converted into a single-path structure ( Figure 6 ), reducing the amount of computation and data storage.

[0076] Preferably, the step of determining the inner arc edge defect area of ​​the curved screen to be inspected based on all microcrack defect detection results includes:

[0077] The inner arc edge defect area of ​​the curved screen to be inspected is determined according to the location of the microcrack defect feature corresponding to each microcrack defect detection result.

[0078] Specifically, the location of the microcrack defect feature corresponding to each microcrack defect detection result is obtained, and the defect area at the location of each microcrack defect feature is extracted and spliced ​​to obtain the inner arc edge defect area of ​​the curved screen to be detected.

[0079] Preferably, it also includes:

[0080] When the area of ​​the inner arc edge defect region is larger than a preset defect area, the curved screen to be inspected is determined to be unqualified and an alarm message is output.

[0081] It should be noted that ① the preset defect area is set based on actual needs and is not limited here. ② In addition to comparing the area of ​​the inner arc edge defect area, it can also be compared based on features such as the defect length, width, and contrast extracted by machine vision blob analysis. If the corresponding threshold is exceeded, the curved screen to be inspected is judged as unqualified and an alarm message is output.

[0082] The technical solution of this embodiment uses half-pupil focusing technology to achieve rapid focusing of the microscope camera, focusing the lens on the TFT circuit layer within the curved screen. This blurs surface interference to distinguish interference from microcracks, providing a more accurate image for subsequent defect detection. This technical solution not only improves the accuracy of microcrack defect detection on the curved edge of the curved screen, but also enhances detection efficiency.

[0083] In the second embodiment of the method for detecting micro crack defects on the inner arc edge of a curved screen provided by the present invention, a detection device that integrates the method for detecting micro crack defects on the inner arc edge of a curved screen provided by the present invention is used for the detection. The detection device adopts a dual-station design, using two detection channels for detection respectively, and splitting the long side and short side detection into two units, so that the device can detect 4 curved screens at the same time, ensuring that the detection device meets the production cycle. Figure 9 As shown, the detection equipment includes: a feeding unit, a positioning unit, a short side detection unit, a screen rotation unit, a long side detection unit and a discharge unit. Specifically:

[0084] ① Feeding unit: used to receive two curved screens to be inspected at the same time and transport them to the alignment unit.

[0085] ② Alignment Unit: Before each curved screen to be inspected is placed into the inspection channel, it must pass through the alignment unit to ensure that the subsequent microscope camera can capture the correct area. The alignment unit uses two separate high-precision visual UVW alignment platforms to correct the screen's position within a ±10 micron error.

[0086] ③ The short side detection unit and the long side detection unit implement the specific manner in the first embodiment of the method for detecting micro-crack defects on the inner arc edge of the curved screen provided by the present invention, which will not be elaborated here.

[0087] ④ Screen rotation unit: used to rotate the curved screen to be tested 90 degrees to facilitate the detection of the long side of the curved screen to be tested.

[0088] ⑤Discharging unit: used to discharge the curved screen to be inspected after the inspection is completed.

[0089] The testing equipment adopts an assembly line design. While the short sides of two curved screens to be tested are being tested, the long sides of the next two curved screens to be tested will also be tested to ensure a production cycle of 4.5s for each curved screen to be tested.

[0090] The detection equipment used in the technical solution of this embodiment solves the problem that the inner arc edge of the curved screen can only be detected by manual sampling. It realizes automatic detection of microcrack defects on the inner arc edge of the curved screen and improves the efficiency of microcrack detection on the inner arc edge of the curved screen.

[0091] Figure 10 FIG. 1 shows a schematic structural diagram of an embodiment of a system for detecting micro crack defects on an inner arc edge of a curved screen provided by the present invention. Figure 10 As shown, the system 200 includes: an acquisition module 210 , a detection module 220 and an operation module 230 .

[0092] The acquisition module 210 is used to: acquire multiple inner arc edge images corresponding to each inner arc edge of the curved screen to be detected; wherein the multiple inner arc edge images corresponding to any inner arc edge include a complete image of the inner arc edge;

[0093] The detection module 220 is used to: input each inner arc edge image into a target defect detection model for detecting microcrack areas, and obtain a microcrack defect detection result of each inner arc edge image;

[0094] The operation module 230 is configured to determine the inner arc edge defect area of ​​the curved screen to be inspected based on all microcrack defect detection results.

[0095] Preferably, the acquisition module 210 is specifically configured to:

[0096] A plurality of inner arc edge images corresponding to each inner arc edge are acquired by using a microscope camera with a half-pupil focusing technique.

[0097] The technical solution of this embodiment uses half-pupil focusing technology to achieve rapid focusing of the microscope camera, focusing the lens on the TFT circuit layer within the curved screen. This blurs surface interference to distinguish interference from microcracks, providing a more accurate image for subsequent defect detection. This technical solution not only improves the accuracy of microcrack defect detection on the curved edge of the curved screen, but also enhances detection efficiency.

[0098] The above-mentioned parameters and steps for each module to implement corresponding functions in the embodiment of the micro-crack defect detection system 200 for the inner arc edge of the curved screen provided by the present invention can be referred to the parameters and steps in the embodiment of the micro-crack defect detection method for the inner arc edge of the curved screen provided above, and will not be repeated here.

[0099] A storage medium provided by an embodiment of the present invention includes: instructions stored in the storage medium, and when a computer reads the instructions, the computer executes the steps of a method for detecting microcrack defects in the inner arc edge of a curved screen. For details, please refer to the parameters and steps in the embodiment of the method for detecting microcrack defects in the inner arc edge of a curved screen provided above, which will not be repeated here.

[0100] Computer storage media such as USB flash drives, mobile hard drives, etc.

[0101] Those skilled in the art will appreciate that the present invention can be implemented as a method, a system, and a storage medium.

[0102] Therefore, the present invention may be embodied in the following forms: entirely in hardware, entirely in software (including firmware, resident software, microcode, etc.), or in a combination of hardware and software, generally referred to herein as a "circuit," "module," or "system." Furthermore, in some embodiments, the present invention may be embodied in the form of a computer program product embodied in one or more computer-readable media, the computer-readable media containing computer-readable program code. Any combination of one or more computer-readable media may be employed. The computer-readable medium may be a computer-readable signal medium or a computer-readable storage medium. A computer-readable storage medium may be, for example, but not limited to, an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device, or any combination thereof. More specific examples (a non-exhaustive list) of computer-readable storage media include: an electrical connection having one or more wires, a portable computer disk, a hard disk, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination thereof. In this document, a computer-readable storage medium may be any tangible medium containing or storing a program that can be used by or in conjunction with an instruction execution system, apparatus, or device. Although embodiments of the present invention have been shown and described above, it should be understood that the above embodiments are exemplary and are not intended to limit the present invention. Those skilled in the art may make changes, modifications, substitutions, and variations to the above embodiments within the scope of the present invention.

Claims

1. A method for detecting micro-crack defects on the inner arc edge of a curved screen, characterized in that: include: Acquire multiple inner arc edge images corresponding to each inner arc edge of the curved screen to be inspected; wherein the multiple inner arc edge images corresponding to any inner arc edge include a complete image of the inner arc edge; Inputting each inner arc edge image into a target defect detection model for detecting microcrack areas, respectively, to obtain a microcrack defect detection result for each inner arc edge image; Based on all microcrack defect detection results, determining the inner arc edge defect area of ​​the curved screen to be detected; The target defect detection model includes: a target image segmentation module, a target image classification module, and a feature extraction module that are sequentially connected and trained; the step of inputting each inner arc edge image into the target defect detection model for detecting the microcrack area to obtain the microcrack defect detection result of each inner arc edge image includes: Input any inner arc edge image into the target image segmentation module, obtain image segmentation features of the inner arc edge image and input them into the target image classification module, obtain image classification features of the inner arc edge image and input them into the feature extraction module, obtain a microcrack defect detection result including the microcrack defect features of the inner arc edge image, until a microcrack defect detection result for each inner arc edge image is obtained; The target image segmentation module includes: a first semantic feature extraction layer, a second semantic feature extraction layer, a third semantic feature extraction layer, a fourth semantic feature extraction layer, a fifth semantic feature extraction layer, a sixth semantic feature extraction layer, a fusion layer, and a seghead layer, which are arranged in sequence, and a first detail feature extraction layer, a second detail feature extraction layer, and a third detail feature extraction layer, which are arranged in sequence. The step of inputting any inner arc edge image into the target image segmentation module to obtain the image segmentation features of the inner arc edge image includes: Inputting any inner arc edge image into the first semantic feature extraction layer and sequentially performing semantic feature extraction on the second semantic feature extraction layer and the third semantic feature extraction layer to obtain a first intermediate feature of the inner arc edge image; Inputting the first intermediate feature of any inner arc edge image into the fourth semantic feature extraction layer and the first detail feature extraction layer respectively, so as to sequentially pass the first intermediate feature input into the fourth semantic feature extraction layer through the fifth semantic feature extraction layer and the sixth semantic feature extraction layer for feature extraction to obtain a second intermediate feature, and sequentially pass the first intermediate feature input into the first detail feature extraction layer through the second detail feature extraction layer and the third detail feature extraction layer to obtain a third intermediate feature; wherein, feature fusion is performed respectively between the fourth semantic feature extraction layer and the first detail feature extraction layer, between the fifth semantic feature extraction layer and the second detail feature extraction layer, and between the sixth semantic feature extraction layer and the third detail feature extraction layer; The second intermediate feature and the third intermediate feature of any inner arc edge image are input into the fusion layer for feature fusion to obtain the target fusion feature and input into the seghead layer for image segmentation processing to obtain the image segmentation feature of the inner arc edge image.

2. The method for detecting micro-crack defects on the inner arc edge of a curved screen according to claim 1, characterized in that: The step of obtaining a plurality of inner arc edge images corresponding to each inner arc edge of the curved screen to be inspected comprises: A plurality of inner arc edge images corresponding to each inner arc edge are acquired by using a microscope camera with a half-pupil focusing technique.

3. The method for detecting micro-crack defects on the inner arc edge of a curved screen according to claim 1, characterized in that: The target image classification module includes: sequentially connecting a plurality of re-parameterized residual structures; inputting the image segmentation features of any inner arc edge image into the target image classification module to obtain the image classification features of the inner arc edge image, including: The image segmentation features of any inner arc edge image are input into the first re-parameterized residual structure and image classification processing is performed in turn through each re-parameterized residual structure to obtain the image classification features of the inner arc edge image.

4. The method for detecting micro-crack defects on the inner arc edge of a curved screen according to claim 1, characterized in that: The step of determining the inner arc edge defect area of ​​the curved screen to be inspected based on all microcrack defect detection results includes: The inner arc edge defect area of ​​the curved screen to be inspected is determined according to the location of the microcrack defect feature corresponding to each microcrack defect detection result.

5. The method for detecting micro-crack defects on the inner arc edge of a curved screen according to any one of claims 1 to 4, characterized in that: Also includes: When the area of ​​the inner arc edge defect region is larger than a preset defect area, the curved screen to be inspected is determined to be unqualified and an alarm message is output.

6. A micro-crack defect detection system for the inner arc edge of a curved screen, characterized in that: include: Get module, detect module and run module; The acquisition module is used to: acquire multiple inner arc edge images corresponding to each inner arc edge of the curved screen to be detected; wherein the multiple inner arc edge images corresponding to any inner arc edge include a complete image of the inner arc edge; The detection module is used to: input each inner arc edge image into a target defect detection model for detecting microcrack areas, and obtain a microcrack defect detection result of each inner arc edge image; The operation module is used to: determine the inner arc edge defect area of ​​the curved screen to be inspected based on all microcrack defect detection results; The target defect detection model includes: a target image segmentation module, a target image classification module, and a feature extraction module that are sequentially connected and trained; the detection module is specifically used to: Input any inner arc edge image into the target image segmentation module, obtain image segmentation features of the inner arc edge image and input them into the target image classification module, obtain image classification features of the inner arc edge image and input them into the feature extraction module, obtain a microcrack defect detection result including the microcrack defect features of the inner arc edge image, until a microcrack defect detection result for each inner arc edge image is obtained; The target image segmentation module includes: a first semantic feature extraction layer, a second semantic feature extraction layer, a third semantic feature extraction layer, a fourth semantic feature extraction layer, a fifth semantic feature extraction layer, a sixth semantic feature extraction layer, a fusion layer and a seghead layer, and a first detail feature extraction layer, a second detail feature extraction layer and a third detail feature extraction layer, which are arranged in sequence; the detection module is specifically used to: Inputting any inner arc edge image into the first semantic feature extraction layer and sequentially performing semantic feature extraction on the second semantic feature extraction layer and the third semantic feature extraction layer to obtain a first intermediate feature of the inner arc edge image; Inputting the first intermediate feature of any inner arc edge image into the fourth semantic feature extraction layer and the first detail feature extraction layer respectively, so as to sequentially pass the first intermediate feature input into the fourth semantic feature extraction layer through the fifth semantic feature extraction layer and the sixth semantic feature extraction layer for feature extraction to obtain a second intermediate feature, and sequentially pass the first intermediate feature input into the first detail feature extraction layer through the second detail feature extraction layer and the third detail feature extraction layer to obtain a third intermediate feature; wherein, feature fusion is performed respectively between the fourth semantic feature extraction layer and the first detail feature extraction layer, between the fifth semantic feature extraction layer and the second detail feature extraction layer, and between the sixth semantic feature extraction layer and the third detail feature extraction layer; The second intermediate feature and the third intermediate feature of any inner arc edge image are input into the fusion layer for feature fusion to obtain the target fusion feature and input into the seghead layer for image segmentation processing to obtain the image segmentation feature of the inner arc edge image.

7. The micro-crack defect detection system for the inner arc edge of a curved screen according to claim 6, characterized in that: The acquisition module is specifically used for: A plurality of inner arc edge images corresponding to each inner arc edge are acquired by using a microscope camera with a half-pupil focusing technique.

8. A storage medium, characterized in that: The storage medium stores instructions, and when a computer reads the instructions, the computer is caused to execute the method for detecting micro-crack defects in the inner arc edge of a curved screen according to any one of claims 1 to 5.

Images