Corner point detection method, system and equipment

Abstract

The invention belongs to the technical field of image processing, particularly relates to a corner detection method, system and equipment, and aims to solve the problem of insufficient precision of a corner detection method in the related technology. The method comprises the steps of obtaining a target image; performing corner detection on the target image through a corner detection model to obtain a primary corner image, and performing mask generation on the target image to obtain a target mask; the corner detection model is constructed based on a U-net structure, the U-net structure comprises an encoder and a decoder, and the decoder comprises a corner detection branch and a mask generation branch; and filtering the primary corner image through the target mask to obtain target corner data. According to the method, corner detection is carried out through cooperation of corner acquisition and mask generation, so that the interference of background information is removed while the global information of the image is reserved, and the corner detection precision and speed are improved.

Description

Background Art

[0002] In the field of image processing technology, corner points are one of the features of image recognition, processing and analysis. How to detect corner points quickly and accurately is of great significance.

[0003] In the related art, corner detection is easily interfered by the background image. However, if the background image is removed before corner detection, there will not be enough distinguishing information, resulting in insufficient corner detection accuracy. At the same time, this method is computationally complex and will also lead to the problem of slow corner detection speed. Summary of the invention

[0004] In order to solve the above problems in the prior art, namely, the problems of insufficient accuracy and slow detection speed of the corner point detection method in the related art, the present invention provides a corner point detection method, which comprises: Get the target image; Through the corner detection model, the target image is subjected to corner detection to obtain a primary corner image, and Perform mask generation on the target image to obtain a target mask; The corner detection model is constructed based on a U-net structure, wherein the U-net structure includes an encoder and a decoder, and the decoder includes a corner detection branch and a mask generation branch; The primary corner point image is filtered through the target mask to obtain target corner point data.

[0005] In some preferred embodiments, the corner detection model includes: Multi-stage encoding and decoding module, corner detection head and mask detection head; The multi-level encoding and decoding module, each level of the encoding and decoding unit includes 1 encoding unit and 1 decoding unit; The output end of the first stage decoding unit is connected to the corner detection head and the mask detection head.

[0006] In some preferred embodiments, the corner detection model is trained by the following method: Prepare training set images under various conditions and from various perspectives; The primary corner point images and masks of the training set are obtained respectively through the untrained corner point detection model; The first loss is calculated by the primary corner image of the training set, and the second loss is calculated by the mask of the training set; Calculate the second target corner point data of the training set by adding the training set mask, and then calculate the third loss; The model parameters are adjusted until the first loss, the second loss and the third loss are all lower than the set threshold, and a trained corner detection model is obtained. The second target corner point data is the output of the corner detection model.

[0007] In some preferred embodiments, the first loss is a loss function corresponding to a corner detection algorithm.

[0008] In some preferred embodiments, the second loss is a binary cross entropy loss.

[0009] In some preferred embodiments, the third loss for: ; in, is the first loss of the i-th image, represents the mask of the i-th image, is the second loss of the i-th image, Represents the set bias term, and N represents the number of images in each group of training set images.

[0010] Another aspect of the present invention provides a corner point detection system, the system comprising: An acquisition unit, used for acquiring a target image; A processing unit, configured to perform corner detection on a target image using a corner detection model to obtain a primary corner image, and Perform mask generation on the target image to obtain a target mask; The corner detection model is constructed based on a U-net structure, wherein the U-net structure includes an encoder and a decoder, and the decoder includes a corner detection branch and a mask generation branch; The primary corner point image is filtered by the target mask to obtain a target corner point image.

[0011] According to a third aspect of the present invention, an electronic device is provided, comprising: at least one processor; and a memory communicatively connected to at least one of the processors; wherein, The memory stores instructions that can be executed by the processor, and the instructions are used to be executed by the processor to implement the above-mentioned corner point detection method.

[0012] According to a fourth aspect of the present invention, a computer-readable storage medium is provided, wherein the computer-readable storage medium stores computer instructions, and the computer instructions are used to be executed by the computer to implement the above-mentioned corner detection method.

[0013] Beneficial effects of the present invention: The present invention performs corner point detection by coordinating corner point acquisition and mask generation in a corner point detection model, thereby retaining the global information of the image and removing the interference of background information, thereby improving the accuracy and speed of corner point detection. BRIEF DESCRIPTION OF THE DRAWINGS

[0014] Other features, objects and advantages of the present application will become more apparent by reading the detailed description of non-limiting embodiments made with reference to the following drawings: Figure 1 is a flowchart of a corner point detection method in an embodiment of the present invention; Figure 2 is a schematic diagram of the principle of a corner point detection method in an embodiment of the present invention; Figure 3 is a structural diagram of a corner point detection model in an embodiment of the present invention; Figure 4 Schematic diagram of a processing object for a Gaussian blur interference test in an embodiment of the present invention; Figure 5 is a schematic diagram of processing result analysis of a Gaussian blur interference test in an embodiment of the present invention; Figure 6 Schematic diagram of a processing object for a Gaussian noise interference test in an embodiment of the present invention; Figure 7 is a schematic diagram of analyzing the processing results of a Gaussian noise interference test in an embodiment of the present invention; Figure 8 is a schematic diagram of a processing object for a perspective angle interference test in an embodiment of the present invention; Fig. 9 2 is a schematic diagram of analyzing the processing results of perspective angle interference in an embodiment of the present invention. DETAILED DESCRIPTION

[0015] The present application will be further described in detail below in conjunction with the accompanying drawings and embodiments. It is to be understood that the specific embodiments described herein are only used to explain the relevant invention, rather than to limit the invention. It should also be noted that, for ease of description, only the parts related to the relevant invention are shown in the accompanying drawings.

[0016] It should be noted that, in the absence of conflict, the embodiments and features in the embodiments of the present application can be combined with each other. The present application will be described in detail below with reference to the accompanying drawings and in combination with the embodiments.

[0017] In related technologies, corner detection mainly relies on traditional image processing algorithms, such as OpenCV, MATLAB, etc. Although these methods perform well under normal conditions, they are often not robust enough under complex conditions such as illumination changes, occlusion, or image noise.

[0018] To solve these problems, some researchers began to explore deep learning-based methods, using the powerful feature extraction capabilities of neural networks to improve the accuracy and robustness of corner detection.

[0019] SuperPoint is a corner point detection algorithm based on deep learning, which identifies and locates corner points in images by training neural networks. The advantage of SuperPoint is that it can detect corner points quickly and accurately, and has good adaptability to various image transformations. However, SuperPoint is not specially designed for checkerboard corner point detection. If it is directly applied to the detection of checkerboard corner points, the problem that the network cannot distinguish between checkerboard corner points and non-checkerboard corner points may be encountered, and the checkerboard corner points may also be missed. The present invention has made further improvements to the problems existing in SuperPoint, adding a mask branch to the SuperPoint network, adjusting the network structure, and increasing the upsampling link to cope with the situation where the chessboard accounts for a relatively small proportion in the image acquisition field of view, thereby improving the detection accuracy and speed.

[0020] U-net is a deep learning network architecture commonly used for image segmentation. It achieves accurate image segmentation through a symmetrical encoder-decoder structure. The encoder part of U-net gradually downsamples the image and extracts multi-level feature maps, while the decoder part gradually restores the original resolution of the image through upsampling and combines the feature maps of the corresponding encoder layer in each upsampling step to achieve feature fusion and detail recovery.

[0021] In order to more clearly illustrate the corner point detection method of the present invention, Figure 1 and Figure 2 Each step in the embodiment of the present invention is described in detail.

[0022] In one embodiment, in large-angle target images, traditional image processing algorithms face challenges of perspective transformation and image deformation, which may lead to inaccurate extraction of checkerboard corner features. Feature points may become blurred and misplaced, affecting the accuracy of corner detection. In images where the target image occupies a small area of ​​the image, due to the small proportion of the target image, its feature points may not be significant enough in the image and are easily interfered by background noise and other irrelevant features. In this case, traditional image processing algorithms often have difficulty in accurately identifying and locating the corners of the target image, resulting in reduced detection performance.

[0023] The corner point detection method of the first embodiment of the present invention includes steps S11 to S13, each of which is described in detail as follows: Step S11, acquiring a target image; Step S12, performing corner detection on the target image using a corner detection model to obtain a primary corner image, and Perform mask generation on the target image to obtain a target mask; The corner detection model is constructed based on a U-net structure, wherein the U-net structure includes an encoder and a decoder, and the decoder includes a corner detection branch and a mask generation branch; In the embodiments of the present disclosure, the corner detection branch can be performed using any existing corner detection method, such as SuperPoint corner detection, Harris corner detection, Shi-Tomasi corner detection, or FAST corner detection, etc., which is not specifically limited in the present disclosure. The mask generation branch can be performed using any existing image segmentation method, such as FCN, SegNet, PSPNet, etc., which is not specifically limited in the present disclosure.

[0024] Step S13: Filter the primary corner point image by the target mask to obtain target corner point data. The target corner point data includes corresponding spatial coordinate information, which can be further used for camera calibration later.

[0025] The present invention filters the primary corner point image through the target mask, and the operation method is as follows: The target mask is overlaid on the primary corner point image to filter the corner points on the non-calibration plate to ensure that the corner points are on the calibration plate, thereby reducing or preventing the corner points on the non-calibration plate from being misdetected as corner points on the calibration plate, thereby reducing the accuracy of the detection.

[0026] Through the above-mentioned embodiment, by setting a corner detection model with parallel corner detection branches and mask generation branches, the coordination of corner detection and mask is achieved. For the corner data to be extracted in the end, the global information is retained and the interference of background information is eliminated, thereby improving the detection accuracy and speed of the final data. Especially for large-angle viewing angle changes, the attention brought by the data-enhanced neural network and the mask can also effectively extract and identify corner features, and maintain high accuracy even after the image has undergone perspective transformation and deformation. And because the interference of background information is filtered out by the mask, the target object can be accurately identified even when it accounts for a small proportion of the overall target image.

[0027] In one embodiment, in a low-resolution image, the features of the checkerboard corners may become blurred and unclear, making it difficult for traditional image processing algorithms to accurately detect and extract these feature points. Low-resolution images have fewer pixels and less information, which makes it easy for feature points to be obscured by background noise and image compression artifacts. The following embodiments of the present disclosure further illustrate how to improve the accuracy of corner detection by setting the corner detection model structure: In the embodiment of the present disclosure, the corner point detection model includes: Multi-stage encoding and decoding module, corner detection head and mask detection head; Optionally, the multi-level encoding and decoding module adopts a U-net structure with a 4-level encoding and decoding structure; the encoder uses resnet-18 as the backbone network, and the decoder part uses a VGG network.

[0028] The multi-level encoding and decoding module, each level of the encoding and decoding unit includes 1 encoding unit and 1 decoding unit; like Figure 3 As shown, the first level coding and decoding unit includes BasicBlock1_Encoder and BasicBlock1_Decoder, the second, third and fourth level coding and decoding units all include BasicBlock2_Encoder and BasicBlock1_Decoder, and the fourth level coding and decoding units are also connected through BasicBlock2_Encoder, wherein the output end of the encoding unit of each level of coding and decoding unit is connected to the input end of the encoding unit of the next level of coding and decoding unit. The output end of the decoding unit of each level of coding and decoding unit is connected to the input end of the decoding unit of the previous level of coding and decoding unit.

[0029] The output of the first-level decoding unit is connected to the corner detection head and the mask detection head, which are responsible for predicting corner points and removing false detections in non-chessboard areas, respectively.

[0030] Through the above embodiments, by using the U-net structure to build a corner detection model, it is possible to further optimize the corner detection architecture by introducing more efficient upsampling and feature enhancement algorithms, so that it can more effectively identify and extract small-scale features of the checkerboard. By introducing a special checkerboard mask, the present invention can accurately filter and enhance the checkerboard features in the detection process, thereby improving the accuracy and stability of corner detection under low-resolution conditions.

[0031] In one embodiment, the corner detection method of the present disclosure can be applied to camera calibration, multi-camera joint reconstruction, target tracking, 3D reconstruction, image fusion and robot navigation. In order to achieve the above purpose, it is necessary to obtain a corner detection model that can adapt to various task types. The following embodiments of the present disclosure further illustrate how to train the corner detection model in the present disclosure.

[0032] In the embodiment of the present disclosure, the corner point detection model is trained by the following method: Configure the initial parameters of the corner detection model, including batch size, learning rate, detection threshold, NMS parameters, etc. Set the number of training iterations, the interval for recording and saving the model, and the frequency of verifying the model performance. Specify whether to retrain the model and the random seed to ensure the repeatability of the experiment.

[0033] Prepare training set images under various conditions and from various perspectives; Optionally, the training set images include images under various shooting conditions. For example, if 3D reconstruction is required, images of similar objects to be reconstructed under various lighting conditions, occlusion, noise, multiple angles, and multiple distances are required as training set images. Taking the calibration of multiple cameras using a chessboard as an example, it is necessary to obtain chessboard images of multiple distances, multiple poses, multiple lighting conditions, multiple contrasts, multiple image proportions, high resolution, low resolution, and shape variations as training set data.

[0034] Optionally, data augmentation of the training set data can be performed by random resizing, photometric adjustments such as brightness, contrast, salt and pepper noise, Gaussian noise, shadows, motion blur, and perspective transformations such as translation, rotation, scaling, and perspective.

[0035] The primary corner point images and masks of the training set are obtained respectively through the untrained corner point detection model; The first loss is calculated by the primary corner image of the training set, and the second loss is calculated by the mask of the training set; In the embodiment of the present disclosure, the specific method of obtaining the primary corner point image is not limited. Taking the SuperPoint corner point detection method as an example, the first loss is: loss1=BCE(Softmax(output), target); Output represents the output data of corner point detection, and target represents the target data.

[0036] Among them, the classifier function is: ; represents the raw score of the ith position, represents the raw score of the j-th position.

[0037] Binary cross entropy is BCE: ; p and y represent the first and second alternative variables in the loss function, represents the first alternative variable of the i-th image, The second substitution variable representing the i-th image, N represents the number of images, and i represents the i-th image.

[0038] In the disclosed embodiment, the second loss is a binary cross entropy loss.

[0039] The binary cross entropy loss is: loss2=BCE(Sigmoid(output), target); Output represents the mask detection output data, and target represents the target data.

[0040] Among them, the Sigmoid function is: ; Calculate the second target corner point data of the training set by adding the training set mask, and then calculate the third loss; In the embodiment of the present disclosure, the third loss for: ; in, is the first loss of the i-th image, represents the mask of the i-th image, is the second loss of the i-th image, Represents the set bias term, and N represents the number of images in each group of training set images.

[0041] The model parameters are adjusted until the first loss, the second loss and the third loss are all lower than the set threshold, and a trained corner detection model is obtained. The second target corner point data is the output of the corner detection model.

[0042] The post-processing process of extracting corner points from the heat map, applying non-maximum suppression (NMS), and removing boundary points. The post-processing technology improves the accuracy of the detection results through confidence threshold screening, NMS to remove duplicate detections, and boundary point removal.

[0043] Through the above embodiments, by setting different loss functions for corner detection, mask generation and overall recognition, each branch focuses on different tasks under the joint action of the three loss functions, and can promote each other during the training process. The scheme disclosed in the present invention can effectively cope with the scenes of camera calibration, multi-camera joint reconstruction, target tracking, three-dimensional reconstruction, image fusion and robot navigation. Improve the accuracy and automation of corner detection. In terms of data enhancement, the present invention implements a variety of data enhancement techniques, including random size adjustment, photometric adjustment and perspective transformation, etc., to simulate various real-world image changes. These data enhancement strategies enhance the model's adaptability to different conditions and ensure that the model can still maintain good detection performance in a variety of complex environments.

[0044] It is understandable that the embodiments of the present disclosure do not simply introduce additional mask branches to filter out background information. More importantly, the coordinated optimization of mask and corner point generation is achieved during the training phase to enhance the corner point detection accuracy for large-angle viewing angles, low resolution, small chessboard proportions, and chessboard images with interference. The following embodiments of the present disclosure further illustrate the application effects of the present disclosure through experimental data and effect demonstration.

[0045] In the disclosed embodiment, Figure 4 is a schematic diagram of a processing object for a Gaussian blur interference test in an embodiment of the present invention, Figure 5 is a schematic diagram of the analysis of the processing results of the Gaussian blur interference test in an embodiment of the present invention; see Figure 4 and Figure 5 Under the same experimental conditions, as the degree of Gaussian blur increases, the corner detection accuracy of the method proposed by the present invention is always higher than that of OpenCV.

[0046] Figure 6 is a schematic diagram of a processing object for a Gaussian noise interference test in an embodiment of the present invention, Figure 7 is a schematic diagram of the analysis of the processing results of the Gaussian noise interference test in an embodiment of the present invention. Figure 6 and Figure 7 Under the same experimental conditions, as the degree of Gaussian noise increases, the corner detection accuracy of the method proposed by the present invention is always higher than that of OpenCV.

[0047] Figure 8 is a schematic diagram of a processing object for a perspective angle interference test in an embodiment of the present invention, Fig. 9 is a schematic diagram of the analysis of the processing results of perspective angle interference in an embodiment of the present invention, see Figure 8 and Fig. 9 ,Under the same experimental conditions, the corner point detection accuracy of the method proposed in the present invention is always higher than that of OpenCV as the perspective tilt angle increases.,Especially when the perspective tilt angle is greater than 70 degrees, the detection method of OpenCV fails.

[0048] Through the above embodiments, the present disclosure adopts a multi-branch network structure, including a corner detection branch and a checkerboard mask branch, each branch focuses on different tasks and can promote each other during the training process. The corner detection branch focuses on detecting checkerboard corners, while the mask branch improves the overall detection accuracy by filtering out false detections in non-checkerboard areas. The multi-branch network structure enhances the ability of feature extraction, especially for large-angle viewing angles, low resolution, small checkerboard proportions, and checkerboard images with interference, significantly improving the accuracy and robustness of detection.

[0049] The introduction of the checkerboard mask branch solves the problem of low feature point significance and susceptibility to background interference in images with a small proportion of checkerboards by accurately filtering and strengthening checkerboard features. The mask branch can accurately identify and filter out false detections in non-chessboard areas, significantly improving the accuracy and robustness of detection, especially when processing images with a small proportion of checkerboards, ensuring high-precision corner point detection.

[0050] The multi-branch design and the introduction of mask branches enhance the model's adaptability to multi-angle viewpoint changes and complex environments. The network is able to handle large-angle viewpoints, low resolution, small chessboard proportions, and chessboard images with interference. Even after perspective transformation and image deformation, the model can maintain high-accuracy corner detection, showing excellent robustness and adaptability. The improved SuperPoint architecture and specific loss function design improve the accuracy of corner detection. The detection process is further optimized by combining the design of mask branches. The model can accurately detect checkerboard corners, significantly reducing the false detection rate and missed detection rate, ensuring high-precision detection results. Although the various steps in the above embodiment are described in the above order, those skilled in the art can understand that in order to achieve the effect of this embodiment, different steps do not have to be executed in such order, and they can be executed simultaneously (in parallel) or in reverse order. These simple changes are within the scope of protection of the present invention. The corner point detection system of the second embodiment of the present invention includes: An acquisition unit, used for acquiring a target image; A processing unit, configured to perform corner detection on a target image using a corner detection model to obtain a primary corner image, and Perform mask generation on the target image to obtain a target mask; The corner detection model is constructed based on a U-net structure, wherein the U-net structure includes an encoder and a decoder, and the decoder includes a corner detection branch and a mask generation branch; The primary corner point image is filtered by the target mask to obtain a target corner point image.

[0051] Those skilled in the art can clearly understand that, for the convenience and brevity of description, the specific working process and related instructions of the system described above can refer to the corresponding process in the aforementioned method embodiment, and will not be repeated here.

[0052] It should be noted that the corner point detection system provided in the above embodiment is only illustrated by the division of the above functional modules. In practical applications, the above functions can be assigned to different functional modules as needed, that is, the modules or steps in the embodiments of the present invention can be decomposed or combined. For example, the modules in the above embodiments can be combined into one module, or further divided into multiple sub-modules to complete all or part of the functions described above. The names of the modules and steps involved in the embodiments of the present invention are only for distinguishing the modules or steps and are not regarded as improper limitations of the present invention.

[0053] An electronic device according to a third embodiment of the present invention includes: at least one processor; and a memory communicatively connected to at least one of the processors; wherein, The memory stores instructions that can be executed by the processor, and the instructions are used to be executed by the processor to implement the above-mentioned corner point detection method.

[0054] A fourth embodiment of the present invention provides a computer-readable storage medium, wherein the computer-readable storage medium stores computer instructions, and the computer instructions are used to be executed by the computer to implement the above-mentioned corner detection method.

[0055] Those skilled in the art can clearly understand that, for the convenience and brevity of description, the specific working process and related instructions of the storage device and processing device described above can refer to the corresponding process in the aforementioned method embodiment and will not be repeated here.

[0056] Those skilled in the art should be able to appreciate that the modules and method steps of each example described in conjunction with the embodiments disclosed herein can be implemented with electronic hardware, computer software, or a combination of the two, and the programs corresponding to the software modules and method steps can be placed in random access memory (RAM), memory, read-only memory (ROM), electrically programmable ROM, electrically erasable programmable ROM, registers, hard disks, removable disks, CD-ROMs, or any other form of storage medium known in the technical field. In order to clearly illustrate the interchangeability of electronic hardware and software, the composition and steps of each example have been generally described in the above description according to the function. Whether these functions are performed in electronic hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered to exceed the scope of the present invention.

[0057] The terms "first", "second", etc. are used to distinguish similar objects rather than to describe or indicate a particular order or sequence.

[0058] The term "comprise" or any other similar term is intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus / device that includes a list of elements includes not only those elements but also other elements not expressly listed, or also includes elements inherent to such process, method, article, or apparatus / device.

[0059] So far, the technical solutions of the present invention have been described in conjunction with the preferred embodiments shown in the accompanying drawings. However, it is easy for those skilled in the art to understand that the protection scope of the present invention is obviously not limited to these specific embodiments. Without departing from the principle of the present invention, those skilled in the art can make equivalent changes or substitutions to the relevant technical features, and the technical solutions after these changes or substitutions will fall within the protection scope of the present invention.

Claims

1. A corner point detection method, characterized in that: include: Get the target image; Through the corner detection model, the target image is subjected to corner detection to obtain a primary corner image, and Perform mask generation on the target image to obtain a target mask; The corner detection model is constructed based on a U-net structure, wherein the U-net structure includes an encoder and a decoder, and the decoder includes a corner detection branch and a mask generation branch; The primary corner point image is filtered through the target mask to obtain target corner point data.

2. The corner point detection method according to claim 1, characterized in that: The corner point detection model comprises: Multi-stage encoding and decoding module, corner detection head and mask detection head; The multi-level encoding and decoding module, each level of the encoding and decoding unit includes 1 encoding unit and 1 decoding unit; The output end of the first stage decoding unit is connected to the corner detection head and the mask detection head.

3. The corner point detection method according to claim 1, characterized in that: The corner detection model is trained by the following method: Prepare training set images under various conditions and from various perspectives; The primary corner point images and masks of the training set are obtained respectively through the untrained corner point detection model; The first loss is calculated by the primary corner image of the training set, and the second loss is calculated by the mask of the training set; Calculate the second target corner point data of the training set by adding the training set mask, and then calculate the third loss; The model parameters are adjusted until the first loss, the second loss and the third loss are all lower than the set threshold, and a trained corner detection model is obtained. The second target corner point data is the output of the corner detection model.

4. The corner point detection method according to claim 3, characterized in that: The first loss is a loss function corresponding to the corner detection algorithm.

5. The corner point detection method according to claim 3, characterized in that: The second loss is the binary cross entropy loss.

6. The corner point detection method according to claim 3, characterized in that: The third loss for: ； in, is the first loss of the i-th image, represents the mask of the i-th image, is the second loss of the i-th image, Represents the set bias term, and N represents the number of images in each group of training set images.

7. A corner point detection system, characterized in that: include: An acquisition unit, used for acquiring a target image; A processing unit, configured to perform corner detection on a target image using a corner detection model to obtain a primary corner image, and Perform mask generation on the target image to obtain a target mask; The corner detection model is constructed based on a U-net structure, wherein the U-net structure includes an encoder and a decoder, and the decoder includes a corner detection branch and a mask generation branch; The primary corner point image is filtered by the target mask to obtain a target corner point image.

8. An electronic device, characterized in that: include: at least one processor; as well as a memory communicatively connected to at least one of the processors; wherein, The memory stores instructions executable by the processor, and the instructions are used to be executed by the processor to implement the corner point detection method according to any one of claims 1 to 6.

9. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores computer instructions, and the computer instructions are used to be executed by the computer to implement the corner point detection method according to any one of claims 1 to 6.

Images