Image recognition method, device, electronic device and storage medium for multi-image fusion

Abstract

The present invention discloses an image recognition method, device, electronic device and storage medium applied to multi-image fusion, belonging to the field of image recognition technology. The method avoids the process of fusion of images, first extracts image features and matches them; then uses the original image input for target detection, and then compares the target boundary box position obtained after detection with the feature matching position, and leaves the target object after screening; thereby retaining the original feature information of the image to the greatest extent, facilitating the extraction of the original features of the image in the input machine learning model, improving the accuracy of image recognition, and solving the problem that the algorithm for image target extraction is not accurate enough due to the difference in samples between the training image and the image actually used.

Description

Technical Field

[0001] The present invention belongs to the technical field of image recognition, and in particular, relates to an image recognition method, device, electronic equipment and storage medium applied to multi-image fusion. Background Art

[0002] In recent years, deep learning methods have become more mature in the application of target detection and recognition in images, and the accuracy of the algorithm has been greatly improved compared with traditional feature extraction. In remote sensing images and aerial photography applications, cameras often cannot capture all objects at once. They will take a large number of photos of repeated superposition of target objects and perform splicing and fusion processing. The spliced ​​images are then used as input to the deep learning network model. The advantage of this is that one image can complete all target objects, but with the influence of the quality of the captured images and the lens angle, the image effect after image splicing and fusion in many cases is not good, so that it is difficult to ensure the validity of the detection results when such images with poor fusion effects are input into the network for feature extraction. Secondly, since the deep learning algorithm is based on sample training, the images used in training are all single-shot original images. After splicing and fusion, the images are all weighted superposition. This results in differences in sample acquisition between the training images and the images actually used, resulting in the detection results being affected by the splicing and fusion effects. Therefore, this algorithm for extracting targets after fusion has great uncertainty. Summary of the invention

[0003] 、Problems to be solved

[0004] In view of the problem that the existing image recognition uses spliced ​​and fused images as training images, due to the difference in samples between the training images and the images actually used, the algorithm for extracting image targets is not accurate enough, the present invention avoids the process of fusing images, first extracts image features and matches them; then uses the original image input for target detection, and then compares the target boundary box position obtained after detection with the feature matching position, and retains the target object after screening; in this way, the original feature information of the image is retained to the greatest extent, which is convenient for inputting the original features of the image into the machine learning model to extract the original features, thereby improving the accuracy of image recognition.

[0005] 、Technical solutions

[0006] To solve the above problems, the present invention adopts the following technical solutions.

[0007] A first aspect of the present invention provides an image recognition method applied to multi-image fusion, the method comprising:

[0008] S102: Acquire an image to be identified, calibrate the image to be identified into a template image and a target image, and respectively extract feature points of the template image and the target image using a feature extraction algorithm;

[0009] S104: traversing the feature points of the template image, filtering the feature point information by distance, and drawing a feature matching area according to the filtered feature point information;

[0010] S106: Detecting the template image and the target image by using a target detection algorithm to obtain detection frames of the target objects in the template image and the target image;

[0011] S108: Calculate the relative position between the detection frame and the feature matching area, and crop the image to be recognized according to the relative position.

[0012] In some embodiments, before the step of acquiring the image to be identified and marking the image to be identified as a template image and a target image, the step further includes:

[0013] Obtain two images to be identified, and calculate the width and height of the two images to be identified;

[0014] In the horizontal direction, the maximum value of the heights of the two images to be identified is used as the height of the splicing; in the vertical direction, the widths of the two images to be identified are added together as the width of the splicing;

[0015] According to the stitching height and width, an image palette is constructed, and one image is randomly calibrated as the template image and the other as the target image.

[0016] In some embodiments, the step of respectively extracting feature points of the template image and the target image by a feature extraction algorithm comprises:

[0017] The features of the template image and the target image are extracted by the AKAZE algorithm, and the extracted feature point information is stored as a matrix data structure.

[0018] In some embodiments, the step of traversing the feature points of the template image and filtering the feature point information by distance includes:

[0019] Traversing the feature points of the template image by using a proximity algorithm to obtain a close distance and a second close distance;

[0020] The ratio of the close distance to the secondary close distance is calculated, and the feature points whose ratio meets the predetermined numerical range are retained as matching values ​​in the matrix data structure; and the feature points whose ratio does not meet the predetermined numerical range are deleted from the matrix data structure.

[0021] In some embodiments, the step of drawing a feature matching area according to the filtered feature point information includes:

[0022] The pixel coordinates corresponding to the feature points are obtained, and the feature matching area is drawn according to the splicing direction and pixel coordinates of the target image and the template image.

[0023] In some embodiments, calculating the relative position relationship between the detection frame and the feature matching area includes:

[0024] According to the coordinate information of the detection frame and the feature matching area;

[0025] Determine whether the detection frame is in the feature matching area; if not, crop the template image and the target image according to the size of the detection frame;

[0026] When the detection frame overlaps with the feature matching area, detect whether the target object is in the overlapping area; crop the template image and the target image according to the judgment result; wherein the detection frame includes the detection frame of the target object in the template image and the target image.

[0027] In some embodiments, when the detection frame overlaps with the feature matching area, detecting whether the target object is within the overlapping area; and the step of cropping the template image and the target image according to the judgment result includes:

[0028] If there is no detected object in the corresponding feature line in the overlapping area, the template image and the target image are cropped according to the size of the detection frame;

[0029] If there is a detected object on the corresponding feature line in the overlap detection, determine whether the object is of the same type as the object in step S106;

[0030] If not, the template image and the target image are correspondingly cropped according to the size of the detection frame;

[0031] If so, crop the detection box with the largest area and output it.

[0032] A second aspect of the present invention provides an image recognition method and device for multi-image fusion, comprising:

[0033] A feature extraction module is used to obtain an image to be identified, mark the image to be identified as a template image and a target image, and respectively extract feature points of the template image and the target image by a feature extraction algorithm;

[0034] A screening module, which is used to traverse the feature points of the template image, screen the feature point information by distance, and draw a feature matching area according to the screened feature point information;

[0035] An object detection module, which is used to detect the template image through an object detection algorithm to obtain a detection frame of the object in the template image; and

[0036] The recognition module is used to calculate the relative position of the detection frame and the feature matching area, and to crop the template image and the target image according to the result of the relative position.

[0037] A third aspect of the present invention provides an electronic device, comprising a processor, an input device, an output device and a memory, wherein the processor, input device, output device and memory are connected in sequence, the memory is used to store a computer program, the computer program comprises program instructions, and the processor is configured to call the program instructions to execute the above method.

[0038] A fourth aspect of the present invention provides a readable storage medium, wherein the storage medium stores a computer program, wherein the computer program includes program instructions, and when the program instructions are executed by a processor, the processor executes the above method.

[0039] , beneficial effects

[0040] Compared with the prior art, the present invention has the following beneficial effects:

[0041] (1) The present invention avoids the process of image fusion, first extracts image features and matches them; then uses the original image input for target detection, and then compares the target bounding box position obtained after detection with the feature matching position, and leaves the target object after screening; thereby retaining the original feature information of the image to the greatest extent, making it easier to extract the original features of the image when input into the machine learning model, thereby improving the accuracy of image recognition;

[0042] (2) The error matching elimination algorithm used in the present invention obtains the ratio of the close distance and the sub-close distance through the KNN algorithm for screening, which can ensure the accuracy of the image feature points to the greatest extent;

[0043] (3) The present invention abandons the fusion algorithm with a long detection time, making the image detection speed faster. BRIEF DESCRIPTION OF THE DRAWINGS

[0044] The above and other purposes, features and advantages of the present application will become more apparent by describing the embodiments of the present application in more detail in conjunction with the accompanying drawings. The accompanying drawings are used to provide a further understanding of the embodiments of the present application and constitute a part of the specification. Together with the embodiments of the present application, they are used to explain the present application and do not constitute a limitation of the present application. In the accompanying drawings, the same reference numerals generally represent the same components or steps. In the accompanying drawings:

[0045] Figure 1 is a flow chart of an image recognition method applied to multi-image fusion provided by an embodiment of the present invention;

[0046] Figure 2 It is a block diagram of an adaptive image splicing and fusion device provided by an embodiment of the present invention;

[0047] Figure 3 is a flow chart of an image adaptive boundary fusion method provided by an embodiment of the present invention;

[0048] Figure 4 is a block diagram of an electronic device provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0049] Below, the embodiments according to the present application will be described in detail with reference to the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present application, rather than all the embodiments of the present application, and it should be understood that the present application is not limited to the embodiments described here.

[0050] Exemplary Methods

[0051] like Figure 1 and 3 As shown, an image recognition method applied to multi-image fusion includes the following steps:

[0052] S102: Acquire an image to be identified, calibrate the image to be identified into a template image and a target image, and respectively extract feature points of the template image and the target image using a feature extraction algorithm.

[0053] Specifically, the image to be identified in this example can obtain the original image of the pointer code table to be identified in a photo or video, for example, read the image through a camera, and use the camera read API to directly read the frame of the current image stream; randomly select an image as a template image, and the remaining images are target images; extract the features of the template image and the target image through the AKAZE algorithm, and store the extracted feature point information as a matrix data structure, where the feature point information includes the feature point information of the template image and the feature point information of the target image. Those skilled in the art should understand that the feature extraction algorithm here can also be a SIFT (Scale Invariant Feature Transform) algorithm, etc., which is not limited here.

[0054] As a variation, before the step of obtaining the image to be identified and marking the image to be identified as a template image and a target image, the method further includes:

[0055] Obtain two images to be identified, and calculate the width and height of the two images to be identified;

[0056] In the horizontal direction, the maximum value of the heights of the two images to be identified is used as the height of the splicing; in the vertical direction, the widths of the two images to be identified are added together as the width of the splicing;

[0057] According to the stitching height and width, an image palette is constructed, and one image is randomly calibrated as the template image and the other as the target image.

[0058] It should be noted that, in the embodiments of the present invention, generally, a template image can be used as a template, and features characterizing the template image or the target of interest therein (referred to as "template features" in this article") can be obtained from the template image. These template features can be points and / or edges that can characterize the template image or the target image; features are extracted from the target image or the template image, and the target features can also be points and / or edges; this example extracts feature point information, but is not limited to this.

[0059] S104: traversing the feature points of the template image, filtering the feature point information by distance, and drawing a feature matching area according to the filtered feature point information.

[0060] Specifically, the feature points of the template image are traversed through the proximity algorithm to obtain the close distance and the second close distance; the feature points here refer to a large number of feature points that can be obtained using the AKAZE detection algorithm, and these feature points will be described by some numerical values ​​(such as illumination, angle, grayscale value, etc.). The close distance and the second close distance are used to illustrate the similarity of KNN screening feature points, using the absolute value of the difference between the feature points of the two images used as the distance (because the calculation method and the distance formula are very similar, it is generally called the feature distance). Since the KNN algorithm provides multiple similarities, the similarities will be sorted according to the distance calculation, and the minimum feature distance value of each feature is selected as the close distance, and the second minimum value is selected as the second close distance.

[0061] The ratio of the close distance to the secondary close distance is calculated, and the close distance points whose ratio meets the predetermined value range are retained as matching values ​​in the matrix data structure; and the close distance points whose ratio does not meet the predetermined value range are deleted from the matrix data structure.

[0062] In this example, the threshold is defined as thread, the close distance value is defined as, and the second close distance value is defined as, calculated according to the method; if, the feature point is retained as a matching point in the matrix data structure determined in step S102; if, the feature point is deleted from the matrix data structure determined in step S102, in this example, the selection of the threshold thread can be determined according to actual conditions and is not limited here. The error matching elimination algorithm used in this example uses the KNN algorithm to obtain the ratio of the close distance and the second close distance for screening, which can maximize the accuracy of the image feature points.

[0063] The pixel coordinates corresponding to the feature points are obtained, and the feature matching area is drawn according to the splicing direction and pixel coordinates of the target image and the template image.

[0064] It should be noted that the feature points detected in this example have pixel coordinates. When horizontal splicing is required, this series of coordinates is sorted according to the x-axis (horizontal direction of the image); if vertical splicing is required, it is sorted according to the y-axis (vertical reverse direction of the image); the sorting method here is to use bubble sort to find the maximum value of the y-axis or x-axis coordinates. If the two images are spliced ​​horizontally from left to right, then the x-axis (horizontal pixels of the image) is used as the separation area; similarly, if the top and bottom are spliced ​​vertically, the y-axis (vertical pixels of the image) is used as the separation area.

[0065] In a possible implementation, the maximum coordinate position of the target image (x1, y1) and the minimum coordinate position of the source image (x2, y2), where x1 is the horizontal coordinate of the maximum coordinate position of the target image, x2 is the horizontal coordinate of the minimum coordinate position of the target image, y1 is the vertical coordinate of the maximum coordinate position of the target image; y2 is the vertical coordinate of the maximum coordinate position of the target image; use them as the two vertex coordinates of the feature matching area, and then use (x1, y2) and (x2, y1) as the other two vertices to obtain a rectangular feature matching area.

[0066] S106: Detect the template image and the target image using a target detection algorithm to obtain detection frames of the target objects in the template image and the target image.

[0067] Specifically, the target detection algorithm here can be to train the target detection YOLO model using the target image training set to obtain a target detection model, and then use the target detection model to detect the test samples to be identified to obtain the coordinate position of the rectangular bounding box of the target object in the test set.

[0068] S108: Calculate the relative position between the detection frame and the feature matching area, and crop the image to be recognized according to the result of the relative position.

[0069] Specifically, according to the coordinates of the detection frame and the feature matching area, it is determined whether the detection frame is in the feature matching area; if not, the target image and the template image are cropped according to the size of the detection frame; wherein the detection frame includes the detection frames of the target objects in the template image and the target image.

[0070] When the detection frame overlaps with the feature matching area, detecting whether the target object is within the overlapping area; and cropping the target image and the template image according to the judgment result;

[0071] If there is no detected object in the corresponding feature line in the overlapping area, the target image and the template image are described according to the size of the detection frame;

[0072] If there is a detected target object on the corresponding feature line in the overlapping detection, determine whether the target object is of the same category as the target object in step S106; the one category here means the class category in target detection. For example, if the detected categories are both people, then a logical judgment is made as to whether there is a repeated area; if one of the detected categories is a person and the other is a vase, then both the person and the vase need to be retained.

[0073] If not, the target image and template image are cropped according to the size of the detection frame; if so, the detection frame with the largest area is cropped and output as the original image. It should be noted that the largest area here refers to the detection frame with the largest area compared with the detection frame in the template image. The one category here means the class category in target detection. For example, if the detected categories are both people, then a logical judgment is made to determine whether there is a repeated area; if one of the detection categories is a person and the other is a vase, then both the person and the vase need to be retained.

[0074] Exemplary Devices

[0075] like Figure 2 As shown, an adaptive image stitching and fusion device comprises:

[0076] The feature extraction module 20 is used to obtain an image to be identified, mark the image to be identified as a template image and a target image, and respectively extract feature points of the template image and the target image by a feature extraction algorithm.

[0077] Specifically, this module is further used to obtain two images to be identified, calculate the width and height of the two images to be identified; in the horizontal direction, use the maximum value of the height of the two images to be identified as the height of the splicing; in the vertical direction, add the widths of the two images to be identified as the width of the splicing; construct an image drawing board according to the height and width of the splicing, and randomly calibrate one image as a template image and the other as a target image; extract the features of the template image and the target image through the AKAZE algorithm, and store the extracted feature point information as matrix data structures respectively.

[0078] The screening module 30 is used to traverse the feature points of the template image, screen the feature point information by distance, and draw the feature matching area according to the screened feature point information. Specifically, the feature points of the template image are traversed by a proximity algorithm to obtain a close distance and a second close distance; the ratio of the close distance and the second close distance is calculated, and the close distance points whose ratio meets the predetermined numerical range are retained in the matrix data structure as matching values; the close distance points whose ratio does not meet the predetermined numerical range are deleted from the matrix data structure; the pixel coordinates corresponding to the feature points are obtained, and the feature matching area is drawn according to the splicing direction of the target image and the template image and the pixel coordinates.

[0079] An object detection module 40 is used to detect the template image by using an object detection algorithm to obtain a detection frame of the object in the template image; and

[0080] The recognition module 50 is used to calculate the relative position between the detection frame and the feature matching area, and to crop the image to be recognized according to the result of the relative position. Specifically, according to the coordinates of the detection frame and the feature matching area; determine whether the detection frame is in the feature matching area; if not, crop the template image and the target image according to the size of the detection frame; when the detection frame overlaps with the feature matching area, detect whether the target object is in the overlapping area; and crop the template image and the target image according to the judgment result. If there is no detected target object on the corresponding feature line in the overlapping area, crop the template image and the target image according to the size of the detection frame; if there is a detected target object on the corresponding feature line in the overlapping detection, determine whether the target object is of the same type as the target object in step S106; if not, crop the template image and the target image according to the size of the detection frame; if so, crop the detection frame with the largest area and output it as the original image.

[0081] Exemplary Electronic Devices

[0082] Below, reference Figure 4 The electronic device according to the embodiment of the present application is described. The electronic device may be the mobile device itself, or a stand-alone device independent of the mobile device, which can communicate with the mobile device to receive the collected input signals from the mobile device and send the combined image information to the mobile device.

[0083] Figure 4 A block diagram of an electronic device according to an embodiment of the present application is illustrated.

[0084] like Figure 4 As shown, the electronic device 10 includes one or more processors 11 and a memory 12 .

[0085] The processor 11 may be a central processing unit (CPU) or other forms of processing units having data processing capabilities and / or instruction execution capabilities, and may control other components in the electronic device 10 to perform desired functions.

[0086] The memory 12 may include one or more computer program products, and the computer program product may include various forms of computer-readable storage media, such as volatile memory and / or non-volatile memory. The volatile memory may include, for example, a random access memory (RAM) and / or a cache memory (cache), etc. The non-volatile memory may include, for example, a read-only memory (ROM), a hard disk, a flash memory, etc. One or more computer program instructions may be stored on the computer-readable storage medium, and the processor 11 may execute the program instructions to implement the decision-making method of each embodiment of the present application described above and / or other desired functions.

[0087] In one example, the electronic device 10 may further include: an input device 13 and an output device 14, which are interconnected via a bus system and / or other forms of connection mechanisms (not shown). For example, the input device 13 may include various devices such as a camera, a video player, etc. The input device 13 may also include, for example, a keyboard, a mouse, etc. The output device 14 may include, for example, a display, a speaker, a printer, a communication network and a remote output device connected thereto, etc.

[0088] Of course, to simplify, Figure 4 Only some of the components related to the present application in the electronic device 10 are shown, and components such as a bus, an input / output interface, etc. are omitted. In addition, according to specific application situations, the electronic device 10 may also include any other appropriate components.

[0089] Exemplary computer program products and computer-readable storage media

[0090] In addition to the above-mentioned methods and devices, an embodiment of the present application may also be a computer program product, which includes computer program instructions, which, when executed by a processor, enable the processor to execute the steps of the decision-making method according to various embodiments of the present application described in the above-mentioned "Exemplary Method" section of this specification.

[0091] The computer program product may be written in any combination of one or more programming languages ​​to write program codes for performing the operations of the embodiments of the present application, including object-oriented programming languages, such as Java, C++, etc., and conventional procedural programming languages, such as "C" language or similar programming languages. The program code may be executed entirely on the user computing device, partially on the user device, as an independent software package, partially on the user computing device and partially on a remote computing device, or entirely on a remote computing device or server.

[0092] In addition, an embodiment of the present application may also be a computer-readable storage medium on which computer program instructions are stored. When the computer program instructions are executed by a processor, the processor executes the steps of the decision-making method according to various embodiments of the present application described in the above "Exemplary Method" section of this specification.

[0093] The computer readable storage medium can adopt any combination of one or more readable media. The readable medium can be a readable signal medium or a readable storage medium. The readable storage medium can include, for example, but is not limited to, a system, device or device of electricity, magnetism, light, electromagnetic, infrared, or semiconductor, or any combination of the above. More specific examples (non-exhaustive list) of readable storage media include: an electrical connection with one or more wires, a portable disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the above.

[0094] The basic principles of the present application are described above in conjunction with specific embodiments. However, it should be noted that the advantages, strengths, effects, etc. mentioned in the present application are only examples and not limitations, and it cannot be considered that these advantages, strengths, effects, etc. are required by each embodiment of the present application. In addition, the specific details disclosed above are only for the purpose of illustration and ease of understanding, not for limitation, and the above details do not limit the present application to being implemented by adopting the above specific details.

[0095] The block diagrams of the devices, apparatuses, equipment, and systems involved in this application are only illustrative examples and are not intended to require or imply that they must be connected, arranged, and configured in the manner shown in the block diagram. As will be appreciated by those skilled in the art, these devices, apparatuses, equipment, and systems can be connected, arranged, and configured in any manner. Words such as "including", "comprising", "having", etc. are open words, referring to "including but not limited to", and can be used interchangeably with them. The words "or" and "and" used here refer to the words "and / or" and can be used interchangeably with them, unless the context clearly indicates otherwise. The words "such as" used here refer to the phrase "such as but not limited to", and can be used interchangeably with them.

[0096] It should also be noted that in the apparatus, device and method of the present application, each component or each step can be decomposed and / or recombined. Such decomposition and / or recombination should be regarded as equivalent solutions of the present application.

[0097] The above description of the disclosed aspects is provided to enable any person skilled in the art to make or use the present application. Various modifications to these aspects will be readily apparent to those skilled in the art, and the general principles defined herein may be applied to other aspects without departing from the scope of the present application. Therefore, the present application is not intended to be limited to the aspects shown herein, but rather to the widest scope consistent with the principles and novel features disclosed herein.

[0098] The above description has been given for the purpose of illustration and description. In addition, this description is not intended to limit the embodiments of the present application to the forms disclosed herein. Although multiple example aspects and embodiments have been discussed above, those skilled in the art will recognize certain variations, modifications, changes, additions and sub-combinations thereof.

Claims

1. An image recognition method applied to multi-image fusion, characterized in that: The method comprises: S102: Acquire an image to be identified, calibrate the image to be identified into a template image and a target image, and respectively extract feature points of the template image and the target image using a feature extraction algorithm; S104: traversing the feature points of the template image, filtering the feature point information by distance, and drawing a feature matching area according to the filtered feature point information; S106: Detecting the template image and the target image by using a target detection algorithm to obtain detection frames of the target objects in the template image and the target image; S108: Calculating the relative position between the detection frame and the feature matching area, and cropping the image to be recognized according to the relative position; Calculating the relative position relationship between the detection frame and the feature matching area includes: According to the coordinate information of the detection frame and the feature matching area; Determine whether the detection frame is in the feature matching area; if not, crop the template image and the target image according to the size of the detection frame; When the detection frame overlaps with the feature matching area, detecting whether the target object is in the overlapping area; cropping the template image and the target image according to the judgment result; wherein the detection frame includes the detection frame of the target object in the template image and the target image; When the detection frame overlaps with the feature matching area, detecting whether the target object is within the overlapping area; and the step of cropping the template image and the target image according to the judgment result includes: If there is no detected object in the corresponding feature line in the overlapping area, the template image and the target image are cropped according to the size of the detection frame; If there is a detected target object on the corresponding feature line in the overlapping area, determine whether the target object is of the same type as the target object in step S106; If not, the template image and the target image are correspondingly cropped according to the size of the detection frame; If so, crop the detection box with the largest area and output it.

2. The image recognition method for multi-image fusion according to claim 1, characterized in that: Before the step of obtaining the image to be identified and marking the image to be identified as a template image and a target image, the method further includes: Obtain two images to be identified, and calculate the width and height of the two images to be identified; In the horizontal direction, the maximum value of the heights of the two images to be identified is used as the height of the splicing; in the vertical direction, the widths of the two images to be identified are added together as the width of the splicing; According to the stitching height and width, an image palette is constructed, and one image is randomly calibrated as the template image and the other as the target image.

3. The image recognition method for multi-image fusion according to claim 2, characterized in that: The step of respectively extracting feature points of the template image and the target image by a feature extraction algorithm comprises: The features of the template image and the target image are extracted by the AKAZE algorithm, and the extracted feature point information is stored as a matrix data structure.

4. The image recognition method for multi-image fusion according to claim 3, characterized in that: The step of traversing the feature points of the template image and filtering the feature point information by distance comprises: Traversing the feature points of the template image by using a proximity algorithm to obtain a close distance and a second close distance; The ratio of the close distance to the secondary close distance is calculated, and the feature points whose ratio meets the predetermined numerical range are retained as matching values ​​in the matrix data structure; and the feature points whose ratio does not meet the predetermined numerical range are deleted from the matrix data structure.

5. The image recognition method for multi-image fusion according to claim 4, characterized in that: The step of drawing the feature matching area according to the filtered feature point information includes: The pixel coordinates corresponding to the feature points are obtained, and the feature matching area is drawn according to the splicing direction and pixel coordinates of the target image and the template image.

6. An image recognition device for multi-image fusion, characterized in that: include: A feature extraction module is used to obtain an image to be identified, mark the image to be identified as a template image and a target image, and respectively extract feature points of the template image and the target image by a feature extraction algorithm; A screening module, which is used to traverse the feature points of the template image, screen the feature point information by distance, and draw a feature matching area according to the screened feature point information; An object detection module is used to detect the template image through an object detection algorithm to obtain a detection frame of the object in the template image; and A recognition module, which is used to calculate the relative position of the detection frame and the feature matching area, and crop the template image and the target image according to the result of the relative position; Calculating the relative position relationship between the detection frame and the feature matching area includes: According to the coordinate information of the detection frame and the feature matching area; Determine whether the detection frame is in the feature matching area; if not, crop the template image and the target image according to the size of the detection frame; When the detection frame overlaps with the feature matching area, detecting whether the target object is in the overlapping area; cropping the template image and the target image according to the judgment result; wherein the detection frame includes the detection frame of the target object in the template image and the target image; When the detection frame overlaps with the feature matching area, detecting whether the target object is within the overlapping area; and the step of cropping the template image and the target image according to the judgment result includes: If there is no detected object in the corresponding feature line in the overlapping area, the template image and the target image are cropped according to the size of the detection frame; If there is a detected target object on the corresponding feature line in the overlapping area, determine whether the target object is of the same type as the target objects in the template image and the target image; If not, the template image and the target image are correspondingly cropped according to the size of the detection frame; If so, crop the detection box with the largest area and output it.

7. An electronic device, characterized in that: The method comprises a processor, an input device, an output device and a memory, wherein the processor, the input device, the output device and the memory are connected in sequence, the memory is used to store a computer program, the computer program comprises program instructions, and the processor is configured to call the program instructions to execute the method according to any one of claims 1 to 5.

8. A readable storage medium, characterized in that: The storage medium stores a computer program, wherein the computer program includes program instructions, and when the program instructions are executed by a processor, the processor executes the method according to any one of claims 1 to 5.

Images