Image element detection method based on corner point information, and network attached storage device

Abstract

The present invention relates to the technical field of image element detection. Disclosed are an image element detection method based on corner point information, and a network attached storage device. The method comprises: on the basis of all training images, performing model training on an image element detection model to be trained, so as obtain a trained image element detection model; and acquiring predicted corner point information corresponding to each training image in the model training process, and on the basis of the predicted corner point information corresponding to all the training images, determining whether the trained image element detection model has converged, and if the trained image element detection model has converged, determining a target image element detection model on the basis of the trained image element detection model, wherein the target image element detection model is used for performing an image element detection operation on an image to be subjected to detection. It can be seen that by means of implementing the present invention, image element detection can be performed by means of a trained target image element detection model, and therefore the reliability and accuracy of image element detection are improved, thereby improving the efficiency of extracting image elements subsequently, and thus facilitating input processing.

Description

Image element detection method based on corner information and network attached storage device Technical Field

[0001] The present invention relates to the technical field of image element detection, and in particular to an image element detection method based on corner point information and a network attached storage device. Background Art

[0002] In scenarios such as real-person authentication and document digitization, it's often necessary to automatically extract document information, such as ID cards, passports, and driver's licenses, for subsequent data entry. However, currently, during document information detection, not only is the detection system susceptible to background interference when identifying document types, but also, due to varying camera angles, documents can easily fail to meet detection requirements, causing text distortion and low text recognition accuracy, thus affecting the proper extraction of document information. Therefore, it's crucial to provide a method that can improve the accuracy of image-based document detection.

[0003] Summary of the Invention

[0004] The present invention provides an image element detection method based on corner point information and a network attached storage device, which improves the reliability and accuracy of image element detection, thereby improving the efficiency of subsequent extraction of image elements for easy input processing.

[0005] In order to solve the above technical problems, the first aspect of the present invention discloses a method for detecting image elements based on corner information, the method comprising:

[0006] Acquire a training image set for training; the training image set includes multiple training images;

[0007] Performing a model training operation on a predetermined image element detection model to be trained based on all the training images to obtain a trained image element detection model;

[0008] Obtain the predicted corner point information corresponding to each of the training images during the model training process, and judge whether the trained image element detection model has converged based on the predicted corner point information corresponding to all the training images. If so, determine the target image element detection model based on the trained image element detection model; the target image element detection model is used to perform image element detection operations on the image to be detected.

[0009] A second aspect of the present invention discloses a network attached storage device, the network attached storage device comprising:

[0010] a memory storing executable program code;

[0011] a processor coupled to the memory;

[0012] The processor calls the executable program code stored in the memory to execute the image element detection method based on corner point information disclosed in the first aspect of the present invention.

[0013] The third aspect of the present invention discloses a computer storage medium, which stores computer instructions. When the computer instructions are called, they are used to execute the image element detection method based on corner point information disclosed in the first aspect of the present invention.

[0014] Compared with the prior art, the embodiments of the present invention have the following beneficial effects:

[0015] In an embodiment of the present invention, a model training operation is performed on an image element detection model to be trained based on all training images used for training to obtain a trained image element detection model. Predicted corner point information corresponding to each training image during the model training process is obtained, and based on the predicted corner point information corresponding to all training images, it is determined whether the trained image element detection model has converged. If so, a target image element detection model is determined based on the trained image element detection model. This target image element detection model is used to perform image element detection on the image to be detected. It can be seen that the implementation of the present invention can perform image element detection using the trained target image element detection model, thereby improving the reliability and accuracy of image element detection, thereby improving the efficiency of subsequent image element extraction, and facilitating input processing. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the following briefly introduces the drawings required for use in the description of the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0017] FIG1 is a schematic flow chart of a method for detecting image elements based on corner information disclosed in an embodiment of the present invention;

[0018] FIG2 is a flow chart of another method for detecting image elements based on corner information disclosed in an embodiment of the present invention;

[0019] FIG3 is a schematic diagram of the structure of a network attached storage device disclosed in an embodiment of the present invention. DETAILED DESCRIPTION

[0020] The present invention discloses an image element detection method based on corner point information and a network attached storage device, which improves the reliability and accuracy of image element detection, thereby improving the efficiency of subsequent extraction of image elements for easy input processing.

[0021] Example 1

[0022] Please refer to Figure 1, which is a flow chart of an image element detection method based on corner information disclosed in an embodiment of the present invention. The image element detection method based on corner information described in Figure 1 can be applied to detect various types of image elements, such as documents, product labels, book covers, etc. in images, which is not limited by the embodiment of the present invention. Optionally, the method can be implemented by an image element detection device, which can be integrated into an image element detection device, such as a smart phone, a smart camera, a smart computer, etc., and when the image element detection device exists independently, it can also be a local server or cloud server for processing the training process of the target image element detection model, which is not limited by the embodiment of the present invention. As shown in Figure 1, the image element detection method based on corner information can include the following operations:

[0023] 101. Obtain a training image set for training.

[0024] In the embodiment of the present invention, the training image set includes a plurality of training images. Optionally, the training images may include RGB training images or grayscale training images.

[0025] 102. Perform a model training operation on a predetermined image element detection model to be trained based on all training images to obtain a trained image element detection model.

[0026] In an embodiment of the present invention, the image element detection model to be trained is used to perform feature merging, binarization, and corner information prediction operations on each training image to achieve the training process of the image element detection model to be trained.

[0027] 103. Obtain the predicted corner point information corresponding to each training image during the model training process, and determine whether the trained image element detection model has converged based on the predicted corner point information corresponding to all training images. If so, determine the target image element detection model based on the trained image element detection model.

[0028] In an embodiment of the present invention, the target image element detection model is used to perform image element detection on the image to be detected. For example, the target image element detection model determines corner point information of the document region in the image to be detected, thereby implementing the image element-document detection process. Optionally, the predicted corner point information includes predicted corner point location information, predicted corner point number information, predicted corner point connection information, and the like.

[0029] Furthermore, as an optional implementation, the method also includes: when it is determined that the trained image element detection model does not converge, updating the model parameters of the trained image element detection model, and using the updated trained image element detection model as a new image element detection model to be trained, so as to trigger the execution of a model training operation for a predetermined image element detection model to be trained based on all training images to obtain the trained image element detection model, and trigger the execution of an operation to obtain the predicted corner point information corresponding to each training image during the model training process, and determine whether the trained image element detection model converges based on the predicted corner point information corresponding to all training images.

[0030] Furthermore, as an optional implementation, the method further includes:

[0031] After determining the target image element detection model, performing an image element detection operation on the image to be detected using the target image element detection model to obtain image elements in the image to be detected; the image elements have corresponding first corner point information;

[0032] According to the first corner point information corresponding to the image element, an image angle affine transformation operation is performed on the image element to obtain a transformed image element; the transformed image element has corresponding second corner point information;

[0033] According to the second corner point information corresponding to the transformed image element, a background removal operation is performed on the image to be detected to obtain the image element after background removal.

[0034] For example, when the document (i.e., image element) in the detected image to be detected is tilted, the four corner points of the document can be determined according to the target image element detection model, and an affine transformation (such as rotation + scaling) can be performed on the four corner points of the document. In this way, the tilted document can be converted into a horizontal and vertical document, and then the blank area can be removed to obtain the final document result.

[0035] It can be seen that the implementation of the embodiment of the present invention can perform image element detection through the trained target image element detection model, thereby improving the reliability and accuracy of image element detection, thereby improving the subsequent efficiency of image element extraction for easy input processing.

[0036] Example 2

[0037] Please refer to Figure 2, which is a flow chart of another method for detecting image elements based on corner information disclosed in an embodiment of the present invention. The method for detecting image elements based on corner information described in Figure 2 can be applied to detect various types of image elements, such as documents, product labels, book covers, etc. in images, which is not limited by the embodiment of the present invention. Optionally, the method can be implemented by an image element detection device, which can be integrated into an image element detection device, such as a smart phone, a smart camera, a smart computer, etc., and when the image element detection device exists independently, it can also be a local server or cloud server for processing the training process of the target image element detection model, which is not limited by the embodiment of the present invention. As shown in Figure 2, the method for detecting image elements based on corner information can include the following operations:

[0038] 201. Obtain a training image set for training.

[0039] 202. Perform a feature merging operation on each training image through a predetermined feature merging layer in the image element detection model to be trained and based on all training images to obtain a feature-merged image corresponding to each training image.

[0040] In an embodiment of the present invention, a feature merging layer (which may be composed of multiple convolutional layers) is used to perform image transformation, feature accumulation merging, channel feature merging and other operations on each training image to obtain a feature-merged image corresponding to each training image.

[0041] 203. Through the binarization processing layer in the image element detection model to be trained, and based on all the feature merged images, a binarization operation is performed on each feature merged image to obtain an approximate binarized image corresponding to each feature merged image.

[0042] In an embodiment of the present invention, a binarization processing layer is used to perform operations such as pixel element probability prediction, binary prediction, and approximate binary calculation on each feature-merged image, and then an approximate binary image corresponding to each feature-merged image is determined based on the obtained approximate binary parameters of all pixels in each feature-merged image.

[0043] 204. Through the prediction layer in the image element detection model to be trained and based on all the approximately binarized images, a corner point information prediction operation is performed on each approximately binarized image to obtain predicted corner point information corresponding to each approximately binarized image.

[0044] In an embodiment of the present invention, the predicted corner point information corresponding to each approximate binary image includes the predicted corner point information of the image elements in the approximate binary image, such as corner point position information, corner point number information, corner point connection method information (such as a rectangular frame formed by four corner points), etc.

[0045] 205. Obtain the predicted corner point information corresponding to each training image during the model training process, and determine whether the trained image element detection model has converged based on the predicted corner point information corresponding to all training images. If so, determine the target image element detection model based on the trained image element detection model.

[0046] In the embodiment of the present invention, for other descriptions of step 201 and step 205, please refer to the detailed description of step 101 and step 103 in embodiment 1, and the embodiment of the present invention will not be repeated.

[0047] It can be seen that the implementation of the embodiment of the present invention can perform feature merging, binarization processing and corner information prediction operations on the training image through the feature merging layer, binarization processing layer and prediction layer in the image element detection model to be trained, and obtain the predicted corner information corresponding to the training image, thereby realizing the training process of the image element detection model to be trained. In this way, the positioning and recognition accuracy of the image elements in the training image can be improved by predicting the corner information, and then the ability of the image elements to be trained to understand the basic geometric structure of the image elements can be improved, thereby improving the training reliability, accuracy and efficiency of the image element detection model to be trained.

[0048] In an optional embodiment, the step 202 above performs a feature merging operation on each training image through a predetermined feature merging layer in the image element detection model to be trained, and based on all training images, obtains a feature-merged image corresponding to each training image, including:

[0049] All training images are input into the feature merging layer of the predetermined image element detection model to be trained, so that the feature merging layer performs the following operations:

[0050] For each training image, performing multiple image transformation operations on the training image according to a preset first image transformation parameter to obtain multiple transformed images corresponding to the training image, and determining, from all the transformed images corresponding to the training image, multiple feature images to be merged and a first image corresponding to each feature image to be merged for feature accumulation and merging;

[0051] For each feature image to be merged corresponding to each training image, performing an image transformation operation on the first image corresponding to the feature image to be merged according to image specification parameters of the feature image to be merged to obtain a transformed first image corresponding to the feature image to be merged, and performing a feature accumulation and merging operation on the feature image to be merged and the transformed first image to obtain an accumulated merged image;

[0052] For each training image, a second image for channel feature merging is determined from all transformed images corresponding to the training image, and an image transformation operation is performed on the second image and all accumulated merged images corresponding to the training image according to preset second image transformation parameters to obtain multiple channel images to be merged corresponding to the training image;

[0053] For each training image, a channel feature merging operation is performed on all channel images to be merged corresponding to the training image to obtain a feature-merged image corresponding to the training image.

[0054] In this optional embodiment, for example, after performing multiple image transformation operations on the training image, 1 / 2 image, 1 / 4 image, 1 / 8 image, 1 / 16 image and 1 / 32 image corresponding to the training image are obtained, and then the 1 / 4 image, 1 / 8 image and 1 / 16 image are used as feature images to be merged, and the 1 / 8 image is used as the first image for feature accumulation merging corresponding to the 1 / 4 image, the 1 / 16 image is used as the first image corresponding to the 1 / 8 image, and the 1 / 32 image is used as the first image corresponding to the 1 / 16 image; then, since the feature accumulation merging operation requires two images of the same size, it is necessary to perform image transformation on the 1 / 8 image, 1 / 16 image and 1 / 32 image (i.e., each first image) in parallel according to the image size parameters of the corresponding feature images to be merged, so as to connect the transformed 1 / 8 image (the transformed 1 / 8 image is equal to the size of the 1 / 4 image, and so on) with the 1 / 4 image, the transformed 1 / 16 image and the original The 1 / 8 image, the transformed 1 / 32 image and the original 1 / 16 image are subjected to feature accumulation and merging operations to obtain three accumulated merged images (with sizes of 1 / 4, 1 / 8 and 1 / 16 respectively); then, the original 1 / 32 image is used as the second image for channel feature merging, and the original 1 / 32 image and the three accumulated merged images are simultaneously transformed into four channel merged images of 1 / 4 size, so that the four channel merged images are merged on the channel layer to obtain a four-channel feature merged image.

[0055] It can be seen that this optional embodiment can perform image transformation, image feature accumulation merging and channel feature merging operations on the training image through the feature merging layer to obtain a feature-merged image corresponding to the training image. In this way, the resistance of the image element detection model to be trained to image changes and noise of the training image can be improved, and the ability of the image element detection model to be trained to capture the feature information in the training image can be improved, thereby improving the reliability and accuracy of the analysis of the image elements contained in the training image, so as to achieve accurate detection of image elements.

[0056] In another optional embodiment, the step 203 above performs a binarization operation on each feature-merged image based on all feature-merged images by using the binarization layer in the image element detection model to be trained, to obtain an approximate binarized image corresponding to each feature-merged image, including:

[0057] The image after all features are merged is input into the binarization processing layer in the image element detection model to be trained, so that the binarization processing layer performs the following operations:

[0058] For each feature-merged image, perform element probability prediction on all pixels in the feature-merged image to obtain the element probability parameter corresponding to each pixel in the feature-merged image;

[0059] For each feature-merged image, a binary prediction operation is performed on all pixels in the feature-merged image to obtain a predicted binary parameter corresponding to each pixel in the feature-merged image;

[0060] For each pixel in the image after each feature is merged, the approximate binary parameter corresponding to the pixel is calculated based on the element probability parameter corresponding to the pixel and the corresponding predicted binary parameter;

[0061] For each feature-merged image, an approximate binary image corresponding to the feature-merged image is determined according to the approximate binary parameters corresponding to all pixels.

[0062] In this optional embodiment, further, the approximate binary parameter corresponding to the pixel is:

[0063] Among them, k is the preset magnification coefficient, P is the element probability parameter corresponding to the pixel, and T is the predicted binary parameter corresponding to the pixel.

[0064] Furthermore, the element probability parameter is used to indicate the predicted probability parameter of the pixel belonging to the image element region. It should be noted that by calculating the approximate binary parameter of the pixel, the output end of the binarization processing layer can be back-propagated.

[0065] It can be seen that this optional embodiment can determine the element probability parameters and predicted binary parameters of each pixel in the image after feature merging through the binarization processing layer, and then calculate the approximate binary parameters of each pixel. In this way, the calculation of the approximate binary parameters can facilitate the subsequent propagation of the gradient of the target loss function to the model parameters, so as to reliably and accurately update the model parameters, and then efficiently train the target image element detection model, so as to accurately detect and extract the image elements in the image to be detected.

[0066] In yet another optional embodiment, the step 205 of determining whether the trained image element detection model has converged based on the predicted corner point information corresponding to all training images includes:

[0067] For each training image, obtaining the region parameters of the marked region in the preset training image, and calculating the region offset of the marked region according to the region parameters of the marked region;

[0068] For each training image, a transition region corresponding to the annotated region is generated according to the region offset of the annotated region, and the corner prediction difference parameter corresponding to the training image is calculated according to the region range information of the transition region and the predicted corner point information corresponding to the training image;

[0069] According to the corner point prediction difference parameters corresponding to all training images, the target loss parameter of the trained image element detection model is calculated, and it is determined whether the target loss parameter is less than or equal to the preset loss parameter. If so, it is determined that the trained image element detection model has converged.

[0070] In this optional embodiment, specifically, the transition area can be understood as the area range that expands inward and outward of the marked area, and after calculating the area offset, when the label value of the marked area is 1, the label values ​​of the areas expanding inward and outward gradually decrease to 0, rather than all being 0, to facilitate model learning.

[0071] Furthermore, the region parameters of the marked region include a region area parameter and a region perimeter parameter of the marked region;

[0072] The area offset of the marked area is:

[0073] Among them, r is the preset hyperparameter (i.e., the range parameter of contraction / expansion), A is the region area parameter, and L is the region perimeter parameter.

[0074] Furthermore, the corner point prediction difference parameter can be understood as the position offset between the predicted corner point of the image element in the training image and the transition area.

[0075] It can be seen that this optional embodiment can calculate the corner point prediction difference parameters corresponding to the training image by predicting the corner point information and the area range information of the transition area corresponding to the marked area, and then perform convergence judgment on the trained image element detection model based on the corner point prediction difference parameters corresponding to each training image. This is conducive to improving the reliability and accuracy of the model's analysis of the position offset relationship between the predicted corner points and the transition area, and thus is conducive to improving the execution reliability and accuracy of the convergence judgment operation of the trained image element detection model, which is conducive to accurately training the target image element detection model that can realize image element detection.

[0076] Example 3

[0077] Please refer to Figure 3, which is a schematic diagram of the structure of a network attached storage device disclosed in an embodiment of the present invention. As shown in Figure 3, the network attached storage device may include:

[0078] A memory 401 storing executable program code;

[0079] a processor 402 coupled to the memory 401;

[0080] The processor 402 calls the executable program code stored in the memory 401 to execute the steps of the image element detection method based on corner point information described in the first embodiment of the present invention or the second embodiment of the present invention.

[0081] Example 4

[0082] An embodiment of the present invention discloses a computer storage medium storing computer instructions. When the computer instructions are called, they are used to execute the steps of the image element detection method based on corner point information described in the first embodiment or the second embodiment of the present invention.

[0083] Through the detailed description of the above embodiments, those skilled in the art can clearly understand that each embodiment can be implemented by means of software plus the necessary general hardware platform, or of course, by means of hardware. Based on this understanding, the above technical solution, in essence, or the portion that contributes to the prior art, can be embodied in the form of a software product, which can be stored in a computer-readable storage medium, including a read-only memory (ROM), a random access memory (RAM), a programmable read-only memory (PROM), an erasable programmable read-only memory (EPROM), a one-time programmable read-only memory (OTPROM), an electronically erasable programmable read-only memory (EEPROM), a compact disc read-only memory (CD-ROM) or other optical disk storage, magnetic disk storage, magnetic tape storage, or any other computer-readable medium capable of carrying or storing data.

Claims

1. A method for detecting image elements based on corner information, characterized in that: The method comprises: Acquire a training image set for training; the training image set includes multiple training images; Performing a model training operation on a predetermined image element detection model to be trained based on all the training images to obtain a trained image element detection model; Obtain the predicted corner point information corresponding to each of the training images during the model training process, and judge whether the trained image element detection model has converged based on the predicted corner point information corresponding to all the training images. If so, determine the target image element detection model based on the trained image element detection model; the target image element detection model is used to perform image element detection operations on the image to be detected.

2. The image element detection method based on corner information according to claim 1, characterized in that: The step of performing a model training operation on a predetermined image element detection model to be trained based on all the training images to obtain a trained image element detection model includes: By using a predetermined feature merging layer in the image element detection model to be trained, and based on all the training images, performing a feature merging operation on each of the training images to obtain a feature-merged image corresponding to each of the training images; Performing a binarization operation on each of the feature-merged images based on all the feature-merged images through the binarization layer in the image element detection model to be trained, to obtain an approximate binarized image corresponding to each of the feature-merged images; Through the prediction layer in the image element detection model to be trained, and based on all the approximate binarized images, a corner point information prediction operation is performed on each of the approximate binarized images to obtain predicted corner point information corresponding to each of the approximate binarized images; the predicted corner point information corresponding to each of the approximate binarized images includes the predicted corner point information of the image elements in the approximate binarized image.

3. The image element detection method based on corner information according to claim 2, characterized in that: The method of performing a feature merging operation on each of the training images through a predetermined feature merging layer in the image element detection model to be trained and based on all the training images to obtain a feature-merged image corresponding to each of the training images includes: All the training images are input into the feature extraction unit of the pre-determined image element detection model to be trained. In the feature merging layer, the feature merging layer performs the following operations: For each of the training images, performing multiple image transformation operations on the training image according to a preset first image transformation parameter to obtain multiple transformed images corresponding to the training image, and determining, from all the transformed images corresponding to the training image, multiple feature images to be merged and a first image corresponding to each of the feature images to be merged for feature accumulation and merging; For each of the feature images to be merged corresponding to each of the training images, performing an image transformation operation on a first image corresponding to the feature image to be merged according to image specification parameters of the feature image to be merged to obtain a transformed first image corresponding to the feature image to be merged, and performing a feature accumulation and merging operation on the feature images to be merged and the transformed first image to obtain an accumulated merged image; For each of the training images, determining a second image for channel feature merging from all the transformed images corresponding to the training image, and performing an image transformation operation on the second image and all the accumulated merged images corresponding to the training image according to preset second image transformation parameters to obtain a plurality of channel images to be merged corresponding to the training image; For each of the training images, a channel feature merging operation is performed on all the channel images to be merged corresponding to the training image to obtain a feature-merged image corresponding to the training image.

4. The image element detection method based on corner information according to claim 2, characterized in that: The binarization layer in the image element detection model to be trained is used, and based on all the feature-merged images, a binarization operation is performed on each of the feature-merged images to obtain an approximate binarized image corresponding to each of the feature-merged images, including: The image after merging all the features is input into the binarization processing layer in the image element detection model to be trained, so that the binarization processing layer performs the following operations: For each of the feature-merged images, performing an element probability prediction operation on all pixels in the feature-merged image to obtain an element probability parameter corresponding to each pixel in the feature-merged image; For each of the feature-merged images, performing a binary prediction operation on all of the pixels in the feature-merged image to obtain a predicted binary parameter corresponding to each of the pixels in the feature-merged image; For each pixel in the image after each feature is merged, the pixel corresponding to the pixel is The pixel probability parameter and the corresponding predicted binary parameter are used to calculate the approximate binary parameter corresponding to the pixel; For each of the feature-merged images, an approximate binary image corresponding to the feature-merged image is determined according to the approximate binary parameters corresponding to all the pixels.

5. The image element detection method based on corner information according to claim 4, characterized in that: The approximate binary parameter corresponding to the pixel is: Wherein, k is a preset magnification coefficient, P is an element probability parameter corresponding to the pixel, and T is a predicted binary parameter corresponding to the pixel.

6. The image element detection method based on corner information according to claim 1, characterized in that: The step of determining whether the trained image element detection model has converged based on the predicted corner point information corresponding to all the training images includes: For each of the training images, obtaining preset regional parameters of the marked region in the training image, and calculating the regional offset of the marked region according to the regional parameters of the marked region; For each of the training images, generating a transition region corresponding to the annotated region according to the region offset of the annotated region, and calculating a corner point prediction difference parameter corresponding to the training image according to region range information of the transition region and predicted corner point information corresponding to the training image; Based on the corner point prediction difference parameters corresponding to all the training images, the target loss parameter of the trained image element detection model is calculated, and it is determined whether the target loss parameter is less than or equal to the preset loss parameter. If so, it is determined that the trained image element detection model has converged.

7. The image element detection method based on corner information according to claim 6, characterized in that: The region parameters of the marked region include a region area parameter and a region perimeter parameter of the marked region; The area offset of the marked area is: Among them, r is a preset hyperparameter, A is the area parameter of the region, and L is the perimeter parameter of the region.

8. The image element detection method based on corner information according to claim 1, characterized in that: The method further comprises: After determining the target image element detection model, performing an image element detection operation on the image to be detected using the target image element detection model to obtain image elements in the image to be detected; the image elements in the image to be detected have corresponding first corner point information; performing an image angle affine transformation operation on image elements in the image to be detected according to the first corner point information to obtain transformed image elements; the transformed image elements have corresponding second corner point information; A background removal operation is performed on the image to be detected according to the second corner point information to obtain image elements after background removal.

9. A network attached storage device, characterized in that: The network attached storage device comprises: a memory storing executable program code; a processor coupled to the memory; The processor calls the executable program code stored in the memory to execute the image element detection method based on corner point information according to any one of claims 1 to 8.

10. A computer storage medium, characterized in that The computer storage medium stores computer instructions, and when the computer instructions are called, they are used to execute the image element detection method based on corner point information according to any one of claims 1 to 8.

Images