Image matching method, system and device based on feature point matching and medium

Abstract

The invention discloses an image matching method, system and device based on feature point matching and a medium, and the method comprises the steps: carrying out the feature point matching of a first image and a second image obtained according to a preset search region, obtaining a plurality of feature point pairs, and filtering the feature point pairs to obtain a first feature point pair set; performing iterative adjustment on the search area according to the first feature point pair set so as to update the first feature point pair set according to the adjusted search area and obtain a second feature point pair set; performing texture analysis on the first image and the second image according to the adjusted search area so as to filter the second feature point pair set and obtain a third feature point pair set; and carrying out global optimization on the third feature point pair set, and taking feature point pairs in the optimized feature point pair set as a matching result of the first image and the second image so as to improve the image matching accuracy.

Description

Technical Field

[0001] The present invention relates to the technical field of image stereo matching, and in particular to an image matching method, system, device and medium based on feature point matching. Background Art

[0002] Stereo vision is an important means for machines to understand the world. The parallax formed by two images can be used to calculate the depth of the pixel points, thereby obtaining three-dimensional stereo perception. Its workflow mainly includes four steps: binocular positioning, image correction, stereo matching and three-dimensional reconstruction, and stereo matching is the key step. The stereo matching algorithm can achieve pixel-level matching of two images. In this process, due to the influence of uneven lighting, occlusion, blur and noise, it brings great challenges to improving the matching accuracy.

[0003] Although binocular stereo matching SLAM technology plays a key role in achieving accurate environmental perception and navigation, it still faces many challenges, especially the problem of low quality of feature point pairs. This is mainly reflected in the algorithm's sensitivity to lighting changes, lack of texture or repeated texture areas, occlusion, and dynamic scenes, resulting in inaccurate matching and insufficient robustness. In addition, the real-time performance and accuracy of existing algorithms when processing large-scale scenes are also limited. Although deep learning methods perform well in some cases, their generalization ability, computing resource consumption, and dependence on specific training data sets are still key bottlenecks in the development of current technology. Therefore, improving the quality of feature point pairs and enhancing the stability and adaptability of the algorithm are the core issues that promote the further development and application of binocular stereo matching SLAM technology. Summary of the invention

[0004] In order to solve the above technical problems, the present invention discloses an image matching method, system, device and medium based on feature point matching, which are used to improve the matching accuracy.

[0005] In order to achieve the above object, in a first aspect, the present invention discloses an image matching method based on feature point matching, comprising:

[0006] Perform feature point matching on the acquired first image and the second image according to a preset search area to obtain a plurality of feature point pairs;

[0007] Filtering the feature point pairs according to preset constraint conditions to obtain a first feature point pair set;

[0008] Iteratively adjusting the search area according to the first feature point pair set, so as to update the first feature point pair set according to the adjusted search area to obtain a second feature point pair set;

[0009] Performing texture analysis on the first image and the second image according to the adjusted search area to filter the second feature point pair set to obtain a third feature point pair set;

[0010] The third feature point pair set is globally optimized, and the feature point pairs in the optimized feature point pair set are used as matching results between the first image and the second image.

[0011] The present invention discloses an image matching method based on feature point matching. First, feature point matching is performed on the acquired first image and the second image according to a preset search area, so as to avoid the extraction of useless feature points by means of the preset search area, reduce the cost of data processing, and thus improve the efficiency of image matching. Then, the feature point pairs are filtered using preset constraints to filter out the feature point pairs that are incorrectly matched, thereby improving the accuracy of feature point matching. Furthermore, in order to improve the accuracy of matching, the search area is iteratively adjusted according to the first feature point pair set, so as to find more matching points in the new area by dynamically adjusting the range of feature point search, thereby improving the accuracy of matching while taking into account the computational cost. Finally, texture analysis and global optimization are performed on the image according to the adjusted search area to improve the stability of matching. Among them, feature point pairs with consistent texture features are selected from the second feature point pair set to improve the accuracy of feature point pair matching.

[0012] As a preferred example, the step of performing feature point matching on the acquired first image and the second image according to the preset search area to obtain a plurality of feature point pairs includes:

[0013] Acquire a plurality of first feature points from the first image according to a preset search area;

[0014] Each of the first feature points and each of the second feature points in the second image are matched according to a preset nearest neighbor search method to obtain a plurality of feature point pairs corresponding to the search area.

[0015] The present invention limits the amount of data to be processed in the initial matching process by presetting the search area, reduces the calculation cost of matching, and improves the matching efficiency. Then, the nearest neighbor search method can be used to quickly find matching feature point pairs, further improving the matching efficiency.

[0016] As a preferred example, filtering the feature point pairs according to the preset constraint conditions to obtain a first feature point pair set includes:

[0017] The feature point pairs are filtered according to a preset geometric consistency constraint algorithm to obtain a first feature point pair set.

[0018] The present invention filters the feature point pairs according to a preset geometric consistency constraint algorithm to screen out incorrectly matched feature point pairs, thereby improving the matching accuracy.

[0019] As a preferred example, the iteratively adjusting the search area according to the first feature point pair set includes:

[0020] Calculating a first distribution center of feature point pairs in the first feature point pair set in the first image and a second distribution center in the second image;

[0021] Calculating, according to the first distribution center and the second distribution center, a first variance of feature point pairs in the first feature point pair set in the first image and a second variance in the second image;

[0022] The search area is iteratively optimized according to the first distribution center, the second distribution center, the first variance, and the second variance until an optimal search area is generated.

[0023] The present invention first dynamically adjusts the range of feature point search according to the current matching situation, so as to find more matching points in the new area, which helps to cover more image areas and thus improve the robustness of matching. Furthermore, it accurately determines the adjustment amount of the search area according to the distribution center and variance of the feature point pairs, so as to reduce the computational cost while reducing the computational cost.

[0024] As a preferred example, the performing texture analysis on the first image and the second image according to the adjusted search area to filter the second feature point pair set includes:

[0025] extracting a first region image from the first image and a second region image from the second image according to the optimal search region;

[0026] Extracting a first local binary pattern feature corresponding to the first region image and a second local binary pattern feature corresponding to the second region image to construct a first histogram corresponding to the first region image and a second histogram corresponding to the second region image;

[0027] The similarity between the first histogram and the second histogram is calculated to filter the second feature point pair set according to the similarity.

[0028] The present invention applies multi-scale and rotation-invariant texture analysis to enhance matching stability and further improve matching accuracy, thereby ensuring the accuracy and consistency of the matching process. In particular, texture analysis is used to select feature point pairs with consistent texture features, thereby improving the matching accuracy of the feature point pairs.

[0029] As a preferred example, the globally optimizing the third feature point pair set includes:

[0030] Acquire the spatial distribution of each feature point pair in the third feature point pair set in the first image to identify a feature point dense area and a feature point sparse area in the first image;

[0031] Extracting local features of the first image in the feature point dense area, and filtering the third feature point pair set according to the local features to obtain a fourth feature point pair set;

[0032] Point cloud matching is performed on the fourth feature point pair set according to a preset iterative closest point algorithm to obtain an optimized feature point pair set.

[0033] The present invention first analyzes the spatial distribution of feature point pairs in the image to identify dense and sparse feature point areas, and then extracts more detailed local features in the dense feature point areas to enhance the accuracy of matching. Finally, the point cloud matching is performed by iterative closest point algorithm to gradually optimize the matching results in an iterative manner, thereby improving the matching accuracy.

[0034] In a second aspect, the present invention discloses an image matching system based on feature point matching, including an initial matching module, a matching filtering module, an iterative updating module, a texture analysis module and a matching optimization module;

[0035] The initial matching module is used to perform feature point matching on the acquired first image and the second image according to a preset search area to obtain a plurality of feature point pairs;

[0036] The matching filtering module is used to filter the feature point pairs according to preset constraint conditions to obtain a first feature point pair set;

[0037] The iterative updating module is used to iteratively adjust the search area according to the first feature point pair set, so as to update the first feature point pair set according to the adjusted search area to obtain a second feature point pair set;

[0038] The texture analysis module is used to perform texture analysis on the first image and the second image according to the adjusted search area to filter the second feature point pair set to obtain a third feature point pair set;

[0039] The matching optimization module is used to perform global optimization on the third feature point pair set, and use the feature point pairs in the optimized feature point pair set as the matching results between the first image and the second image.

[0040] The present invention discloses an image matching system based on feature point matching. First, feature point matching is performed on the acquired first image and the second image according to a preset search area, so as to avoid the extraction of useless feature points by means of the preset search area, reduce the cost of data processing, and thus improve the efficiency of image matching. Then, the feature point pairs are filtered using preset constraints to filter out the feature point pairs that are incorrectly matched, thereby improving the accuracy of feature point matching. Furthermore, in order to improve the accuracy of matching, the search area is iteratively adjusted according to the first feature point pair set, so as to find more matching points in the new area by dynamically adjusting the range of feature point search, thereby improving the accuracy of matching while taking into account the computational cost. Finally, texture analysis and global optimization are performed on the image according to the adjusted search area to improve the stability of matching. Among them, feature point pairs with consistent texture features are selected from the second feature point pair set to improve the accuracy of feature point pair matching.

[0041] As a preferred example, the initial matching module includes a region search unit and a feature point matching unit;

[0042] The area search unit is used to obtain a plurality of first feature points from the first image according to a preset search area;

[0043] The feature point matching unit is used to match each of the first feature points with each of the second feature points in the second image according to a preset nearest neighbor search method to obtain a plurality of feature point pairs corresponding to the search area.

[0044] The present invention limits the amount of data to be processed in the initial matching process by presetting the search area, reduces the calculation cost of matching, and improves the matching efficiency. Then, the nearest neighbor search method can be used to quickly find matching feature point pairs, further improving the matching efficiency.

[0045] In a third aspect, the present invention discloses an electronic device, comprising: at least one processor; and a memory communicatively connected to at least one of the processors; wherein the memory stores instructions executable by at least one of the processors, and the instructions are executed by at least one of the processors so that at least one of the processors can execute an image matching method based on feature point matching as described in the first aspect.

[0046] In a fourth aspect, the present invention discloses a computer-readable storage medium, wherein the computer-readable storage medium stores computer instructions, and the computer instructions are used to enable a computer to execute an image matching method based on feature point matching as described in the first aspect. BRIEF DESCRIPTION OF THE DRAWINGS

[0047] Figure 1: A flowchart of an image matching method based on feature point matching disclosed in an embodiment of the present invention;

[0048] Figure 2 : is a structural schematic diagram of an image matching system based on feature point matching disclosed in an embodiment of the present invention;

[0049] Figure 3 : A flowchart of an image matching method based on feature point matching disclosed in another embodiment of the present invention. DETAILED DESCRIPTION

[0050] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0051] Embodiment 1

[0052] This embodiment discloses an image matching method based on feature point matching. For the specific implementation process of the matching method, please refer to Figure 1 , mainly including steps 101 to 105, wherein the steps are mainly:

[0053] Step 101: performing feature point matching on the acquired first image and the acquired second image according to a preset search area to obtain a plurality of feature point pairs.

[0054] In this embodiment, this step mainly includes: obtaining multiple first feature points from the first image according to a preset search area; matching each of the first feature points with each of the second feature points in the second image according to a preset nearest neighbor search method to obtain multiple feature point pairs corresponding to the search area.

[0055] In this embodiment, this step limits the amount of data to be processed in the initial matching process by presetting the search area, reducing the matching calculation cost to improve the matching efficiency. Then, the nearest neighbor search method can be used to quickly find matching feature point pairs, further improving the matching efficiency.

[0056] Step 102: Filter the feature point pairs according to preset constraints to obtain a first feature point pair set.

[0057] In this embodiment, this step mainly includes: filtering the feature point pairs according to a preset geometric consistency constraint algorithm to obtain a first feature point pair set.

[0058] In this embodiment, this step filters the feature point pairs according to a preset geometric consistency constraint algorithm to filter out incorrectly matched feature point pairs, thereby improving the matching accuracy.

[0059] Step 103: iteratively adjusting the search area according to the first feature point pair set, so as to update the first feature point pair set according to the adjusted search area and obtain a second feature point pair set.

[0060] In this embodiment, this step mainly includes: calculating the first distribution center of the feature point pairs in the first feature point pair set in the first image and the second distribution center in the second image; calculating the first variance of the feature point pairs in the first feature point pair set in the first image and the second variance in the second image based on the first distribution center and the second distribution center; iteratively optimizing the search area based on the first distribution center, the second distribution center, the first variance and the second variance until the optimal search area is generated.

[0061] In this embodiment, this step first dynamically adjusts the range of the feature point search according to the current matching situation, so as to find more matching points in the new area, which helps to cover more image areas, thereby improving the robustness of the matching. Further, it accurately determines the adjustment amount of the search area according to the distribution center and variance of the feature point pairs, so as to reduce the computational cost while reducing the computational cost.

[0062] Step 104: Perform texture analysis on the first image and the second image according to the adjusted search area to filter the second feature point pair set to obtain a third feature point pair set.

[0063] In this embodiment, this step mainly includes: extracting a first area image from the first image and extracting a second area image from the second image according to the optimal search area; extracting a first local binary pattern feature corresponding to the first area image and a second local binary pattern feature corresponding to the second area image to construct a first histogram corresponding to the first area image and a second histogram corresponding to the second area image; calculating the similarity between the first histogram and the second histogram to filter the second feature point pair set according to the similarity.

[0064] In this embodiment, this step applies multi-scale and rotation-invariant texture analysis to enhance matching stability and further improve matching accuracy, thereby ensuring the accuracy and consistency of the matching process. In particular, texture analysis is used to select feature point pairs with consistent texture features, thereby improving the accuracy of matching the feature point pairs.

[0065] Step 105: globally optimizing the third feature point pair set, and using the feature point pairs in the optimized feature point pair set as the matching results of the first image and the second image.

[0066] In this embodiment, this step mainly includes: obtaining the spatial distribution of each feature point pair in the third feature point pair set in the first image to identify feature point dense areas and feature point sparse areas in the first image; extracting local features of the first image in the feature point dense areas to filter the third feature point pair set according to the local features to obtain a fourth feature point pair set; performing point cloud matching on the fourth feature point pair set according to a preset iterative closest point algorithm to obtain an optimized feature point pair set.

[0067] In this embodiment, this step first analyzes the spatial distribution of feature point pairs in the image to identify dense feature point areas and sparse feature point areas, and then extracts more detailed local features in the dense feature point areas to enhance the accuracy of matching. Finally, point cloud matching is performed by iterative closest point algorithm to gradually optimize the matching results in an iterative manner, thereby improving the matching accuracy.

[0068] On the other hand, this embodiment also discloses an image matching system based on feature point matching. For the specific structure of the matching system, please refer to Figure 2 , including an initial matching module 201, a matching filtering module 202, an iterative updating module 203, a texture analysis module 204 and a matching optimization module 205.

[0069] The initial matching module 201 is used to perform feature point matching on the acquired first image and the second image according to a preset search area to obtain a plurality of feature point pairs.

[0070] The matching filtering module 202 is used to filter the feature point pairs according to preset constraint conditions to obtain a first feature point pair set.

[0071] The iterative updating module 203 is used to iteratively adjust the search area according to the first feature point pair set, so as to update the first feature point pair set according to the adjusted search area to obtain a second feature point pair set.

[0072] The texture analysis module 204 is used to perform texture analysis on the first image and the second image according to the adjusted search area to filter the second feature point pair set to obtain a third feature point pair set.

[0073] The matching optimization module 205 is used to perform global optimization on the third feature point pair set, and use the feature point pairs in the optimized feature point pair set as the matching results of the first image and the second image.

[0074] In this embodiment, the initial matching module 201 includes a region search unit and a feature point matching unit.

[0075] The area search unit is used to obtain a plurality of first feature points from the first image according to a preset search area.

[0076] The feature point matching unit is used to match each of the first feature points with each of the second feature points in the second image according to a preset nearest neighbor search method to obtain a plurality of feature point pairs corresponding to the search area.

[0077] According to Figure 1 The image matching method based on feature point matching is shown in the figure. Accordingly, this embodiment provides an electronic device, including a processor, a memory, and a computer program stored in the memory and configured to be executed by the processor. When the processor executes the computer program, the following is implemented Figure 1The image matching method based on feature point matching is shown in the figure. The processor may be a central processing unit (CPU), or other general processors, digital signal processors (DSP), application specific integrated circuits (ASIC), field-programmable gate arrays (FPGA) or other programmable logic devices, discrete gates or transistor logic devices, discrete hardware components, etc. The general processor may be a microprocessor or any conventional processor, etc. The processor is the control center of the terminal, and various interfaces and lines are used to connect various parts of the entire terminal. The memory may be used to store the computer program, and the processor realizes various functions of the terminal by running or executing the computer program stored in the memory and calling the data stored in the memory. The memory may mainly include a program storage area and a data storage area, wherein the program storage area may store an operating system, an application required for at least one function (such as a sound playback function, an image playback function, etc.), etc.; the data storage area may store data created according to the use of the mobile phone (such as audio data, a phone book, etc.), etc. In addition, the memory may include a high-speed random access memory, and may also include a non-volatile memory, such as a hard disk, a memory, a plug-in hard disk, a smart memory card (Smart Med ia Card, SMC), a secure digital (Secure Digital, SD) card, a flash card (FlashCard), at least one disk storage device, a flash memory device, or other volatile solid-state storage devices.

[0078] Furthermore, this embodiment also provides a computer-readable storage medium, wherein the computer-readable storage medium includes a stored computer program, wherein when the computer program is executed, the device where the computer-readable storage medium is located is controlled to execute the following Figure 1 The image matching method based on feature point matching is shown.

[0079] Among them, Figure 2If the image matching system based on feature point matching shown is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, all or part of the processes in the above-mentioned embodiment method can also be completed by instructing the relevant hardware through a computer program, and the computer program can be stored in a computer-readable storage medium. When the computer program is executed by the processor, the steps of the above-mentioned various method embodiments can be implemented. Among them, the computer program includes computer program code, and the computer program code can be in source code form, object code form, executable file or some intermediate form. The computer-readable medium can include: any entity or device that can carry the computer program code, recording medium, U disk, mobile hard disk, disk, optical disk, computer memory, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), electrical carrier signal, telecommunication signal and software distribution medium, etc.

[0080] Compared with the prior art, this embodiment provides the following beneficial effects:

[0081] The present embodiment discloses an image matching method, system, device and medium based on feature point matching. First, feature point matching is performed on the acquired first image and second image according to a preset search area, so as to avoid the extraction of useless feature points by means of the preset search area, reduce the cost of data processing, and thus improve the efficiency of image matching. Then, the feature point pairs are filtered using preset constraints to filter out wrongly matched feature point pairs and improve the accuracy of feature point matching. Furthermore, in order to improve the accuracy of matching, the search area is iteratively adjusted according to the first feature point pair set, so as to find more matching points in the new area by dynamically adjusting the range of feature point search, thereby improving the accuracy of matching while taking into account the computational cost. Finally, texture analysis and global optimization are performed on the image according to the adjusted search area to improve the stability of matching. Among them, feature point pairs with consistent texture features are selected from the second feature point pair set to improve the accuracy of feature point pair matching.

[0082] Embodiment 2

[0083] Stereo vision is an important means for machines to understand the world. Its workflow mainly includes binocular localization, image correction, stereo matching and 3D reconstruction. Among them, stereo matching, as a key step, plays a key role in achieving accurate environmental perception and navigation. In the prior art, when stereo matching is performed on the acquired images, due to the influence of uneven lighting, occlusion, blur and noise, the matching is inaccurate and the robustness is insufficient.

[0084] In response to the above technical problems, this embodiment proposes an image matching method based on an adaptive region growing strategy, which dynamically adjusts the search range according to image texture features and re-evaluates within the optimized range to improve the accuracy of matching point pairs.

[0085] Specifically, the process of the image matching method can be specifically referred to Figure 3 ,like Figure 3 As shown, the matching method includes steps 301 to 305, and the steps are mainly:

[0086] Step 301: extracting a plurality of first feature points from an acquired detection image according to a preset initial search area, and matching the first feature points with each second feature point of a pre-saved reference image to obtain a plurality of feature point pairs.

[0087] In this embodiment, this step mainly includes: first, extracting a plurality of first feature points corresponding to the acquired detection image according to a preset initial search area. Specifically, the initial search area is used to limit the extraction range of the first feature points. In some implementations of this embodiment, the initial search area can be expressed as [x0, y0, w0, h0], where x0 and y0 are the coordinates of the upper left corner of the initial search area, and w0 and h0 are the width and height of the area, respectively.

[0088] While extracting the first feature points, a plurality of second feature points corresponding to the pre-saved reference image for matching are called, and the first feature points and the second feature points are matched one-to-one through a preset feature point matching algorithm to obtain a plurality of feature point pairs.

[0089] Preferably, in some implementations of this embodiment, the feature point matching algorithm is set to use KD-Tree (K-dimensional tree) to match feature points. The KD-Tree can accelerate the feature point matching process, and it helps to quickly find matching feature point pairs through an efficient nearest neighbor search method.

[0090] Specifically, the nearest neighbor search method needs to calculate the KD-Tree distance metric between different feature points, and the calculation expression of the KD-Tree distance metric is:

[0091]

[0092] The two variables p and q represent two points in k-dimensional space. Each point has a coordinate consisting of k values, one for each dimension in k-dimensional space. i ,q i They represent the coordinate values ​​of points p and q in the i-th dimension respectively. The dimension that maximizes the variance is selected for division.

[0093] Step 302: Perform anomaly filtering on the feature point pairs to obtain a first feature point pair set.

[0094] In this embodiment, this step mainly includes: after completing the matching of the first feature point and the second feature point, the initially matched feature point pairs may be screened to remove the erroneous feature point pairs that are mismatched.

[0095] Preferably, in some implementations of this embodiment, a geometric consistency constraint is used to remove incorrect matching point pairs and only retain high-quality matching point pairs. Specifically, the set consistency constraint can be constructed by a RANSAC algorithm. The core formula of the RANSAC algorithm is:

[0096]

[0097] Among them, d(p i ,θ) represents point p i The distance between θ and the model parameter θ, and t is the threshold.

[0098] Step 303: iteratively adjusting the initial search area according to the first feature point pair set, so as to reconstruct the feature point pair set according to the adjusted search area to obtain a second feature point pair set.

[0099] In this embodiment, this step mainly includes: firstly, determining the distribution range of the first feature point in the feature point pair in the detection image and the distribution range of the second feature point pair in the reference image according to the first feature point pair set, so as to adjust the range of the initial search area through the distribution range.

[0100] Specifically, in some implementations of this embodiment, the reference image is set to I1 and the acquired detection image is set to I2, then the feature point set in the reference image is {p1}, and the feature point set in the detection image is {p2}, and a set of matching point pairs {(p1, p2)} is obtained through preliminary matching. In order to dynamically adjust the search area, the distribution center and variance of the matching point pairs can be calculated. The calculation expressions of the distribution center and the variance are as follows:

[0101]

[0102] in, and are the distribution centers of the matching point pairs in the two images, and are the variances of the matching point pairs in the two images. The search area is dynamically adjusted according to the distribution center and variance of the matching point pairs.

[0103] Furthermore, in the process of feature matching, the initial search area is used to limit the range of searching for matching points in the detection image. Based on the assumption that the initial search area is [x0, y0, w0, h0], the process expression for adjusting the initial search area can be:

[0104]

[0105] Among them, k is a tuning parameter that controls the size of the search area.

[0106] Similarly, the search area of ​​the reference image I2 is adjusted accordingly to match new feature point pairs from the reference image and the standard image according to the adjusted search area to add them to the initial feature point pair set, and then the search area adjustment process and the initial feature point pair set update step are repeated to iteratively optimize the search area and feature point pairs until convergence or reaching a predetermined number of iterations, thereby obtaining a second feature point pair set corresponding to the optimal search area.

[0107] Step 304: performing texture analysis on the detection image and the reference image respectively according to the adjusted search area, filtering the second feature point pair set according to the result of the texture analysis, and generating a third feature point pair set.

[0108] In this embodiment, this step mainly includes: analyzing the image texture of the reference image and the standard image within the adjusted search area, so as to further optimize the second feature point pair set through texture analysis. Among them, applying multi-scale and rotation-invariant texture analysis can enhance matching stability, so as to further improve matching accuracy and ensure the accuracy and consistency of the matching process.

[0109] Specifically, in this embodiment, the local binary pattern features corresponding to the reference image and the standard image within the adjusted search area are extracted and a histogram is generated to describe the image texture, and then feature point pairs with consistent texture features are screened out based on the histogram similarity. The texture analysis uses a method based on local binary pattern (LBP) as follows:

[0110]

[0111] Among them, is the sign function of s(x), g k is the surrounding pixel value, g c is the center pixel value.

[0112] Step 305: globally optimizing the third feature point pair set, so as to use the optimized feature point pair set as the matching result between the detection image and the reference image.

[0113] In this embodiment, this step mainly includes: analyzing the spatial distribution of the feature points in the third feature point pair set, identifying areas where the feature points are dense or sparse, and performing global optimization on all feature point pairs.

[0114] Preferably, in certain implementations of this embodiment, the global optimization includes spatial distribution analysis, feature point screening, point cloud matching, multi-scale analysis and post-processing. Specifically, the spatial distribution analysis is to identify the dense feature point areas and sparse feature point areas by analyzing the spatial distribution of feature point pairs in the image. And the feature point screening is to enhance the accuracy of matching by extracting more detailed local features in the dense feature point areas. Next, the point cloud matching is performed using an iterative closest point algorithm, which gradually optimizes the matching results in an iterative manner, thereby improving the matching accuracy. Furthermore, performing multi-scale analysis, that is, performing feature point analysis at different scales, helps to capture features of different sizes and enhance the robustness of matching. After the iterative closest point algorithm is completed, a post-processing step is performed to remove outliers and smoothing to further improve the quality and stability of matching. Among them, the calculation expression of the iterative closest point algorithm is:

[0115] min∑ i ||(Τ(P i )-q i )|| 2

[0116] Wherein, T represents the translation transformation matrix.

[0117] Compared with the prior art, this embodiment provides the following beneficial effects:

[0118] The present embodiment discloses an image matching method based on feature point matching. First, the search range is dynamically adjusted according to the texture features of the image, and the matching point pairs are re-evaluated within the optimized range to effectively deal with problems such as illumination changes, lack of texture, repeated texture areas, occlusion and dynamic scenes, thereby improving the accuracy and robustness of the matching point pairs. In addition, in the preliminary feature point matching stage, KD-Tree is combined for efficient feature point matching, and an adaptive region growth strategy is introduced for further optimization, thereby improving the real-time performance and accuracy when processing large-scale scenes. According to the image matching method provided by the present embodiment, the positioning accuracy and navigation stability of the SLAM system in complex environments can be significantly improved, overcoming the limitations of traditional methods. Further, through texture analysis and feature distribution analysis, the matching method provided by the present embodiment conducts an in-depth analysis of the matching area, further optimizing the quality of the matching point pairs. This not only improves the accuracy of the matching point pairs, but also enhances the stability of the algorithm in complex environments. In addition, after the preliminary matching point pairs are screened, the matching point pairs are re-evaluated and optimized through the adaptive region growth strategy to ensure high-quality matching. This re-evaluation mechanism effectively reduces mismatched points, improves matching accuracy and robustness, and is an innovative design lacking in the prior art.

[0119] The specific embodiments described above further illustrate the purpose, technical solutions and beneficial effects of the present invention. It should be understood that the above description is only a specific embodiment of the present invention and is not intended to limit the scope of protection of the present invention. It is particularly pointed out that for those skilled in the art, any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention should be included in the scope of protection of the present invention.

Claims

1. An image matching method based on feature point matching, characterized in that: include: Perform feature point matching on the acquired first image and the second image according to a preset search area to obtain a plurality of feature point pairs; Filtering the feature point pairs according to preset constraint conditions to obtain a first feature point pair set; Iteratively adjusting the search area according to the first feature point pair set, so as to update the first feature point pair set according to the adjusted search area to obtain a second feature point pair set; Performing texture analysis on the first image and the second image according to the adjusted search area to filter the second feature point pair set to obtain a third feature point pair set; The third feature point pair set is globally optimized, and the feature point pairs in the optimized feature point pair set are used as matching results between the first image and the second image.

2. The image matching method based on feature point matching according to claim 1, characterized in that: The step of performing feature point matching on the acquired first image and the second image according to the preset search area to obtain a plurality of feature point pairs includes: Acquire a plurality of first feature points from the first image according to a preset search area; Each of the first feature points and each of the second feature points in the second image are matched according to a preset nearest neighbor search method to obtain a plurality of feature point pairs corresponding to the search area.

3. The image matching method based on feature point matching according to claim 1, characterized in that: The filtering the feature point pairs according to the preset constraint condition to obtain a first feature point pair set includes: The feature point pairs are filtered according to a preset geometric consistency constraint algorithm to obtain a first feature point pair set.

4. The image matching method based on feature point matching according to claim 2, characterized in that: The iteratively adjusting the search area according to the first feature point pair set includes: Calculating a first distribution center of feature point pairs in the first feature point pair set in the first image and a second distribution center in the second image; Calculating, according to the first distribution center and the second distribution center, a first variance of feature point pairs in the first feature point pair set in the first image and a second variance in the second image; The search area is iteratively optimized according to the first distribution center, the second distribution center, the first variance, and the second variance until an optimal search area is generated.

5. The image matching method based on feature point matching according to claim 4, characterized in that: The performing texture analysis on the first image and the second image according to the adjusted search area to filter the second feature point pair set includes: extracting a first region image from the first image and a second region image from the second image according to the optimal search region; Extracting a first local binary pattern feature corresponding to the first region image and a second local binary pattern feature corresponding to the second region image to construct a first histogram corresponding to the first region image and a second histogram corresponding to the second region image; The similarity between the first histogram and the second histogram is calculated to filter the second feature point pair set according to the similarity.

6. The image matching method based on feature point matching according to claim 5, characterized in that: The globally optimizing the third feature point pair set includes: Acquire the spatial distribution of each feature point pair in the third feature point pair set in the first image to identify a feature point dense area and a feature point sparse area in the first image; Extracting local features of the first image in the feature point dense area, and filtering the third feature point pair set according to the local features to obtain a fourth feature point pair set; Point cloud matching is performed on the fourth feature point pair set according to a preset iterative closest point algorithm to obtain an optimized feature point pair set.

7. An image matching system based on feature point matching, characterized in that: It includes initial matching module, matching filtering module, iterative updating module, texture analysis module and matching optimization module; The initial matching module is used to perform feature point matching on the acquired first image and the second image according to a preset search area to obtain a plurality of feature point pairs; The matching filtering module is used to filter the feature point pairs according to preset constraint conditions to obtain a first feature point pair set; The iterative updating module is used to iteratively adjust the search area according to the first feature point pair set, so as to update the first feature point pair set according to the adjusted search area to obtain a second feature point pair set; The texture analysis module is used to perform texture analysis on the first image and the second image according to the adjusted search area to filter the second feature point pair set to obtain a third feature point pair set; The matching optimization module is used to perform global optimization on the third feature point pair set, and use the feature point pairs in the optimized feature point pair set as the matching results between the first image and the second image.

8. The image matching system based on feature point matching according to claim 7, characterized in that: The initial matching module includes an area search unit and a feature point matching unit; The area search unit is used to obtain a plurality of first feature points from the first image according to a preset search area; The feature point matching unit is used to match each of the first feature points with each of the second feature points in the second image according to a preset nearest neighbor search method to obtain a plurality of feature point pairs corresponding to the search area.

9. An electronic device, characterized in that: include: at least one processor; And a memory communicatively connected to at least one of the processors; wherein the memory stores instructions executable by at least one of the processors, and the instructions are executed by at least one of the processors so that at least one of the processors can execute an image matching method based on feature point matching as described in any one of claims 1-6.

10. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores computer instructions, and the computer instructions are used to enable a computer to execute an image matching method based on feature point matching according to any one of claims 1 to 6.

Images