Feature point connection line-based feature matching error verification method

Verify feature matching by connecting feature points, and use the target matrix and slope judgment to screen out wrong matching, which solves the problem of mismatch in image feature matching and improves the reliability of application scenarios.

CN119006862APending Publication Date: 2024-11-22TIANJIN JINHANG COMP TECH RES INST
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202411187143.8
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-08-28
Publication Date
2024-11-22

AI Technical Summary

Technical Problem

In the prior art, there are a large number of mismatch feature points when matching image features, resulting in matrix solution deviation and camera pose distortion, and error accumulation is difficult to correct, affecting the reliability of application scenarios.

Method used

A feature matching error verification method based on feature point connection lines is proposed. By obtaining image shooting information and feature matching results, the target matrix screening error matching is calculated, and whether the slope of feature point connection lines is within the preset range is confirmed to be matched errors. .

Benefits of technology

Improve the reliability of feature matching, reduce catastrophic errors caused by mismatch, and enhance the reliability of image correction, image mapping and visual robot SLAM systems.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN119006862A_ABST
    Figure CN119006862A_ABST
Patent Text Reader

Abstract

The invention discloses a feature matching error verification method based on a feature point connection line, and relates to the technical field of computer vision and computer images, and the method comprises the steps: obtaining the shooting information of a first image and a second image, and a first feature matching result; based on the shooting information and the first feature matching result, a target matrix is obtained through calculation, so that wrong matching in the first feature matching result is screened out, and a second feature matching result is obtained; and connecting the associated matching feature points in the second feature matching result, judging whether the slope of each connecting line is within a preset range, and if not, determining that the feature points corresponding to the connecting lines are matched wrongly. According to the method, the reliability of feature point matching is improved, and the reliability of application scenes such as subsequent image correction, image mapping and image three-dimensional reconstruction is further improved.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present disclosure generally relates to the field of computer vision and computer image technology, and in particular to a method for verifying feature matching errors based on feature point connection lines. Background Art

[0002] Image feature matching is the process of pixel-level recognition, matching, and alignment of identical or similar feature points in two images. Image scenes are complex and changeable and contain a large number of similar feature points. There will be a large number of mismatched feature points during feature matching. These mismatched feature points will cause a large deviation in the matrix solution, extremely distort the correspondence between images, or cause the camera posture to swing extremely.

[0003] If feature matching errors cannot be identified, an erroneous matrix will be calculated, and the accumulated errors will be difficult to correct, resulting in catastrophic errors in the application scenario.

[0004] Therefore, it is of great significance to verify the mismatch of feature matching and judge the availability of feature matching in advance for application scenarios such as image correction, image mapping, image 3D reconstruction, and visual robot SLAM system. To this end, the present invention proposes a feature matching error verification method based on feature point connection. Summary of the invention

[0005] In view of the above defects or deficiencies in the prior art, it is desirable to provide a feature matching error verification method based on feature point connection lines to improve the reliability of feature point matching.

[0006] The present application provides a method for verifying feature matching errors based on feature point connection lines, comprising:

[0007] Acquire shooting information of the first image and the second image and a first feature matching result;

[0008] Based on the shooting information and the first feature matching result, a target matrix is ​​calculated to filter out wrong matches in the first feature matching result to obtain a second feature matching result;

[0009] Connect the associated matching feature points in the second feature matching result to determine whether the slope of each of the connecting lines is within a preset range. If not, it is determined that the feature point corresponding to the connecting line is incorrectly matched.

[0010] According to the technical solution provided in the present application, obtaining the shooting information of the first image and the second image and the first feature matching result specifically includes:

[0011] receiving the first image and the second image, and extracting matching feature points between the first image and the second image, recording the matching feature points as a first feature matching result;

[0012] Acquire shooting information of the first image and the second image, the shooting information including: a planar image and a non-planar image; in the planar image, the feature points are in the same plane; in the non-planar image, the feature points are not in the same plane.

[0013] According to the technical solution provided in this application, based on the shooting information and the first feature matching result, the target matrix is ​​calculated, specifically including:

[0014] If the first image and the second image are the planar images, a first target matrix is ​​calculated based on the first feature matching result;

[0015] If the first image and the second image are the non-planar images, a second target matrix is ​​calculated based on the first feature matching result.

[0016] According to the technical solution provided by the present application, filtering out the wrong matches in the first feature matching result to obtain the second feature matching result specifically includes:

[0017] When the first image and the second image are the planar images, verifying each matched feature point in the first feature matching result by using the first target matrix;

[0018] Each matching feature point that has not passed the verification of the first target matrix is ​​screened out to obtain the second feature matching result.

[0019] According to the technical solution provided by the present application, the erroneous matches in the first feature matching result are screened out to obtain the second feature matching result, which specifically also includes:

[0020] When the first image and the second image are non-planar images, verifying each matched feature point in the first feature matching result by using the second target matrix;

[0021] Each matching feature point that has not passed the verification of the second target matrix is ​​screened out to obtain the second feature matching result.

[0022] According to the technical solution provided by the present application, determining whether the slope of each of the connecting lines exceeds a preset range specifically includes:

[0023] Calculating the slope corresponding to the line connecting each associated matching point in the second feature matching result and the average of the first slopes corresponding to all matching points;

[0024] Filtering a target slope value set whose difference between the slope of all matching point connecting lines and the first slope mean is within a first preset slope range;

[0025] Based on the target slope set, a second slope mean and a standard deviation are calculated, and the preset range is established with the standard deviation to confirm whether the slope of each of the connecting lines exceeds the preset range.

[0026] According to the technical solution provided by the present application, based on the target slope set, the second slope mean and standard deviation are calculated, and the preset range is established with the standard deviation to confirm whether the slope of each of the connecting lines exceeds the preset range, specifically including:

[0027] Determine whether the difference between the slope corresponding to the connecting line of each associated matching point in the second feature matching result and the second slope average is within the preset range, and if not, confirm that the matching of the feature point corresponding to the connecting line is wrong;

[0028] If yes, it is confirmed that the feature points corresponding to the connecting line are matched correctly.

[0029] According to the technical solution provided in the present application, the first target matrix is ​​a homography matrix calculated based on the matching feature points in the first feature matching result.

[0030] According to the technical solution provided in the present application, the second target matrix is ​​a basic matrix calculated based on the matching feature points in the first feature matching result.

[0031] To summarize, the present technical solution specifically discloses a method for verifying feature matching errors based on connecting feature points, including: obtaining shooting information of a first image and a second image and a first feature matching result; calculating a target matrix based on the shooting information and the first feature matching result to screen out erroneous matches in the first feature matching result to obtain a second feature matching result; connecting the associated matching feature points in the second feature matching result to determine whether the slope of each connecting line is within a preset range, and if not, confirming that the feature point corresponding to the connecting line has a matching error.

[0032] In the existing image feature matching, if the feature matching error cannot be identified, an erroneous matrix will be calculated, and the accumulated errors are difficult to correct, resulting in catastrophic errors in the application scenario. Therefore, it is very important to verify whether the feature point matching is wrong. The present application improves the reliability of image feature point matching by doubly judging whether the feature point is a wrong match based on the homography matrix and the basic matrix and the geometric relationship of the feature point connection line, while also improving the reliability of subsequent image correction, image mapping, image 3D reconstruction, visual robot SLAM system and other application scenarios. BRIEF DESCRIPTION OF THE DRAWINGS

[0033] Other features, objects and advantages of the present application will become more apparent by reading the detailed description of non-limiting embodiments made with reference to the following drawings:

[0034] Figure 1 The figure is a flowchart of a feature matching error verification method based on feature point connection lines.

[0035] Figure 2 A schematic diagram of the principle of a feature matching error verification method based on feature point connection lines.

[0036] Figure 3 Schematic diagram of the expanded flow of step S100 in a feature matching error verification method based on feature point connection lines.

[0037] Figure 4 Schematic diagram of the expanded flow of step S300 in a feature matching error verification method based on feature point connection lines.

[0038] Figure 5 Schematic diagram of the expanded flow of step S303 in a feature matching error verification method based on feature point connection lines.

[0039] Figure 6 Schematic diagram of feature point extraction (the picture comes from the TUM public dataset).

[0040] Figure 7 Based on Figure 6 Schematic diagram of incorrect matching after removing erroneous feature points through the homography matrix (the picture comes from the TUM public dataset).

[0041] Figure 8 Based on Figure 6 Schematic diagram of incorrect matching after removing erroneous feature points through the basic matrix (the picture comes from the TUM public dataset).

[0042] Fig. 9 Schematic diagram of the geometric relationship for correct feature point matching (the image comes from the TUM public dataset).

[0043] Fig.10 for Figure 7 Schematic diagram of abnormal slope of connecting lines caused by incorrect feature matching (the picture comes from TUM public dataset).

[0044] Fig.11 for Figure 8 Schematic diagram of abnormal slope of connecting lines caused by incorrect feature matching (the picture comes from TUM public dataset). DETAILED DESCRIPTION

[0045] The present application is further described in detail below in conjunction with the accompanying drawings and embodiments. It is to be understood that the specific embodiments described herein are only used to explain the relevant invention, rather than to limit the invention. It is also necessary to explain that, for ease of description, only the parts related to the invention are shown in the accompanying drawings.

[0046] It should be noted that, in the absence of conflict, the embodiments and features in the embodiments of the present application can be combined with each other. The present application will be described in detail below with reference to the accompanying drawings and in combination with the embodiments.

[0047] Example 1

[0048] In order to make the technical solution of the embodiment of the present application clearer and easier to understand, the application background of the embodiment of the present application is introduced below.

[0049] Image feature matching is the process of pixel-level recognition, matching, and alignment of identical or similar feature points in two images. Image scenes are complex and changeable and contain a large number of similar feature points. There will be a large number of mismatched feature points during feature matching. These mismatched feature points will cause a large deviation in the matrix solution, extremely distort the correspondence between images, or cause the camera posture to swing extremely.

[0050] If the feature matching errors cannot be identified, the wrong matrix will be calculated, and the accumulated errors are difficult to correct, resulting in catastrophic errors in the application scenario. If the feature matching errors can be identified, the erroneous results will not be used in subsequent applications, and the frame can be discarded for processing.

[0051] Therefore, it is of great significance to verify the false matching of feature matching and judge the availability of feature matching in advance, which is of great significance for application scenarios such as image correction, image mapping, image 3D reconstruction, visual robot SLAM system, etc.

[0052] In existing application scenarios, researchers use cross-check, ratio test and random sampling consistency (RANSAC) methods to remove mismatched feature points. Although these methods can remove a large number of mismatched feature points, they still cannot guarantee the reliability of feature matching.

[0053] In view of this, if Figure 2 As shown, the present invention proposes a feature matching mismatch verification method based on feature point connection lines, which is based on the following inventive concept: ① input two pictures for feature matching to obtain preliminary feature matching results; ② use the random sampling consistency (RANSAC)-homography matrix method to remove the mismatched feature points to obtain the feature matching result; ③ use the basic matrix to remove the mismatched feature points to obtain the feature matching result; ④ match the feature point connection lines, and judge whether the feature matching of ② or ③ is a mismatch based on the geometric relationship of the feature point connection lines.

[0054] Specifically, please refer to Figure 1 The flowchart of a verification method for feature matching errors based on feature point connection lines provided in this embodiment is shown, and the verification method includes the following steps:

[0055] S100, acquiring shooting information of a first image and a second image and a first feature matching result;

[0056] S200, calculating a target matrix based on the shooting information and the first feature matching result to filter out wrong matches in the first feature matching result to obtain a second feature matching result;

[0057] S300, connecting the associated matching feature points in the second feature matching result, and determining whether the slope of each of the connecting lines is within a preset range; if not, confirming that the feature point corresponding to the connecting line is incorrectly matched.

[0058] Exemplarily, the method of the embodiment of the present application may be executed by a terminal, which may be a foldable mobile phone, a tablet computer, a laptop computer, or other device. Specifically, the method includes:

[0059] The terminal uses SIFT feature extraction, SURF feature extraction, ORB feature extraction or AKAZE feature extraction methods to extract matching feature points in the first image and the second image to form a first feature matching result. The first feature matching result here can be understood as a data set or a data list consisting of multiple matching feature point coordinates; and the shooting information is mainly used to distinguish whether the feature points are in the same plane. The shooting information can be defined when the first image and the second image are shot. For example, if the object shot is a piece of paper with content, a book scan, a wall, a desktop or a tablecloth, etc., the feature points are in the same plane; if the object shot is As if it is a real three-dimensional world, the feature points on the picture are larger when they are closer to the camera and smaller when they are farther away, so the feature points are not on the same plane; then the terminal specifically processes the first feature matching result obtained and the picture form corresponding to the shooting information to obtain a target matrix, and then uses the target matrix to screen out incorrect matches in the first feature matching result to obtain a second feature matching result with more accurate matching effect, and finally connects the matching feature points in the second feature matching result to determine whether the slope of each line is within a preset range, and verifies the incorrect matches of the matching feature points in the second feature matching result based on the judgment result, and finally the terminal can obtain the feature matching result with the most accurate matching effect.

[0060] It can be seen that in this method, the terminal performs a secondary screening of the matched feature points, that is, the obtained matched feature points are verified for false matches, which improves the reliability and accuracy of feature matching. In addition, this method combines the features of different images and formulates different screening methods for different image forms, further improving the comprehensiveness and accuracy of false match verification.

[0061] It should be noted that the above is only an example of the method being executed by a terminal. In other examples, the method can also be executed by an application, plug-in, etc. installed in the terminal. This application does not limit the execution subject of the method. However, in order to make the technical solution of this application clearer and easier to understand, the following description is still based on the perspective of the terminal and the server in conjunction with the accompanying drawings.

[0062] In a preferred embodiment, see Figure 3 S100, obtaining shooting information of the first image and the second image and a first feature matching result, further comprising:

[0063] S101, receiving the first image and the second image, and extracting matching feature points between the first image and the second image, which are recorded as first feature matching results;

[0064] Receive two images that need to be compared for feature points, and extract matching feature points in the first image and the second image by using SIFT feature extraction, SURF feature extraction, ORB feature extraction, or AKAZE feature extraction, and record them as first feature matching results;

[0065] Exemplarily, for example, if (x1, y1) in the first image matches (X1, Y1) in the second image, then a set of corresponding matching feature point coordinate data (x1, y1)(X1, Y1) can be obtained, and ultimately multiple sets of matching feature points can be obtained, thereby constituting the first feature matching result.

[0066] S102. Acquire shooting information of the first image and the second image, the shooting information including: a planar image and a non-planar image; in the planar image, the feature points are in the same plane; in the non-planar image, the feature points are not in the same plane.

[0067] The shooting information is mainly used to distinguish whether the feature points are in the same plane. The shooting information can be defined when the first image and the second image are taken. For example, if the photographed object is a piece of paper with content, a book scan, a wall, a desktop or tablecloth, etc., then it is defined as a planar image, and its feature points are in the same plane; if the photographed object is the real three-dimensional world, the feature points on the picture are larger when they are closer to the camera and smaller when they are farther away, then it is defined as a non-planar image, and some of its feature points are not in the same plane.

[0068] In a preferred embodiment, step S200, calculating a target matrix based on the shooting information and the first feature matching result, specifically includes:

[0069] If the first image and the second image are the planar images, a first target matrix is ​​calculated based on the first feature matching result;

[0070] If the first image and the second image are the non-planar images, a second target matrix is ​​calculated based on the first feature matching result.

[0071] like Figure 6 As shown, the first feature matching result includes a large number of correctly matched feature points, and may also include a number of incorrectly matched feature points. Therefore, the incorrectly matched feature points need to be eliminated.

[0072] After the feature points are matched, the correspondence between the feature points can be used to solve the homography matrix (HomographyMatrix) or the fundamental matrix (FundamentalMatrix);

[0073] The homography matrix is ​​used to describe the transformation relationship between two planes and is usually used for perspective transformation between images. The basic matrix describes the relationship between the corresponding epipolar lines in the two images. It assumes that the images are taken from different perspectives by a non-distorted camera, so it is necessary to confirm the calculation of the corresponding target matrix based on different shooting information.

[0074] Among them, the homography matrix represents the pixel correspondence between two images (planes), and the basic matrix represents the rotation and translation of the camera between two images (non-planar features). Homography matrix and basic matrix are widely used in applications such as image correction, image mapping, image 3D reconstruction, and visual robot SLAM system modeling. Therefore, since feature matching is the basis of these application scenarios, the homography matrix and basic matrix are introduced in the embodiment of the present application to verify the mismatch of the matched feature points.

[0075] Exemplarily, here, the first target matrix is ​​a homography matrix calculated based on the matching feature points in the first feature matching result; the second target matrix is ​​a basic matrix calculated based on the matching feature points in the first feature matching result.

[0076] In a preferred embodiment, step S200, filtering out the wrong matches in the first feature matching results to obtain the second feature matching results, specifically includes:

[0077] Step A1: when the first image and the second image are the planar images, verify each matching feature point in the first feature matching result using the first target matrix;

[0078] Step A2: Screen out each matching feature point that has not passed the first target matrix verification to obtain the second feature matching result.

[0079] Specifically, when the feature points are in the same plane, the obtained homography matrix is ​​used to verify the first feature matching result. The verification method is to use the following formula (1): first image feature point coordinates × homography matrix = second image feature point coordinates. The coordinates of each matched feature point are substituted into formula (1) to verify the feature points in the first feature matching result. The feature points that do not satisfy formula (1) are incorrectly matched feature points, and the matched feature points are then removed.

[0080] In some other embodiments, step S200, filtering out the wrong matches in the first feature matching result to obtain the second feature matching result, specifically includes:

[0081] Step B1: when the first image and the second image are non-planar images, using the second target matrix to verify each matching feature point in the first feature matching result;

[0082] Step B2: Screen out each matching feature point that has not passed the second target matrix verification to obtain the second feature matching result.

[0083] Specifically, when the feature point is in a non-plane, the first feature matching result is verified using the obtained basic matrix. The verification method is to use the following formula (2): first image feature point coordinates × basic matrix = second image feature point coordinates. Substitute the coordinates of each matched feature point into formula (2) to verify the matching feature points in the first feature matching result. The feature points that do not satisfy formula (2) are incorrectly matched feature points, and the matching feature points are then removed.

[0084] It should be explained that the use of homography matrix and basic matrix to screen matching feature points is effective in most scenarios. However, if there are many wrong matching feature points in the first feature matching result, resulting in a large deviation between the homography matrix and the basic matrix, it will also cause deviations in the wrong matching verification. For example, Figure 7 and Figure 8 In the case shown, the incorrectly matched feature points still exist.

[0085] In a preferred embodiment, see Figure 4 Step S300, determining whether the slope of each of the connecting lines exceeds a preset range, specifically includes:

[0086] S301, calculating the slope corresponding to the line connecting each associated matching point in the second feature matching result and the average of the first slopes corresponding to all matching points;

[0087] Here we further verify the incorrectly matched feature points based on the geometric relationship between the feature point connections. Fig. 9In the correct feature point matching, the lines between the feature points should be parallel lines or trapezoidal, while the slopes of the lines between the incorrectly matched feature points are inconsistent with those of the correct feature matching lines. Fig.10 and Fig.11 shown Figure 7 and Figure 8 The error matching feature point connection status in .

[0088] The slope of the line connecting the feature matching points can be obtained by calculating the coordinate data of the matching feature points. For example, if the coordinate data of the matching feature points is (x1, y1)(X1, Y1), then the ratio of the ordinate difference to the abscissa difference of the two coordinates can be calculated to obtain the corresponding line slope.

[0089] After obtaining the slopes of all associated matching points, the first slope mean value can be calculated and recorded as M.

[0090] S302, screening a target slope value set whose difference between the slope of all matching point connecting lines and the first slope mean is within a first preset slope range;

[0091] Based on the calculated first slope mean M, the associated matching points in the second feature matching result are screened. The first preset slope range here is optionally set to 80%, without any specific limitation. After the first preset slope range is determined, 80% of the matching feature points whose line slopes of all matching feature points in the second feature matching result have the smallest difference with the first slope mean M constitute the target slope value set Q.

[0092] S303: Based on the target slope set, calculate a second slope mean and a standard deviation, and establish the preset range with the standard deviation to confirm whether the slope of each of the connecting lines exceeds the preset range.

[0093] Based on the matching feature points in the target slope value set Q, the second slope mean M1 and the standard deviation D can be calculated; wherein the standard deviation D can be used as a statistical indicator reflecting the degree of dispersion of a set of measurement data. Here, the preset range established based on the standard deviation can be 0.8D-1.2D, and can also be set according to actual conditions, without specific limitation.

[0094] In a preferred embodiment, see Figure 5 S303: Based on the target slope set, a second slope mean and a standard deviation are calculated, and the preset range is established with the standard deviation to confirm whether the slope of each of the connecting lines exceeds the preset range, specifically including:

[0095] S3031, determining whether the difference between the slope corresponding to the connecting line of each associated matching point in the second feature matching result and the second slope mean is within the preset range, if not, proceeding to S3032, if yes, proceeding to S3033;

[0096] S3032, confirming that the feature point corresponding to the connecting line is matched incorrectly;

[0097] S3033, confirming that the feature points corresponding to the connecting line are matched correctly.

[0098] Perform a final verification on the associated matching points in the second feature matching result, and calculate whether the difference between the slope of the line connecting all the matching feature points in the second feature matching result and the second slope mean M1 is within the range of 0.8D-1.2D. If the difference between the slope of the line connecting all the matching feature points and the second slope mean M1 is less than 0.8D or greater than 1.2D, it is defined as an incorrectly matched feature point.

[0099] Based on the above description, it can be seen that the present application proposes a method for verifying feature matching errors based on connecting feature points, including: obtaining shooting information of a first image and a second image and a first feature matching result; based on the shooting information and the first feature matching result, calculating a corresponding first target matrix or a second target matrix to screen out erroneous matches in the first feature matching result and obtain a second feature matching result; connecting the associated matching feature points in the second feature matching result to determine whether the slope of each connecting line is within a preset range, and if not, confirming that the feature point corresponding to the connecting line is incorrectly matched.

[0100] This method uses the homography matrix, the basic matrix, and the geometric relationship of the feature point connection line to perform dual judgment to verify whether the feature point is a wrong match, which effectively improves the reliability of feature matching. At the same time, it also further improves the reliability of subsequent application scenarios such as image correction, image mapping, image 3D reconstruction, and visual robot SLAM system.

[0101] The above description is only a preferred embodiment of the present application and an explanation of the technical principles used. Those skilled in the art should understand that the scope of the invention involved in the present application is not limited to the technical solution formed by a specific combination of the above technical features, but should also cover other technical solutions formed by any combination of the above technical features or their equivalent features without departing from the inventive concept. For example, the above features are replaced with (but not limited to) technical features with similar functions disclosed in the present application.

Claims

1. A verification method for feature matching errors based on feature point connection lines, It is characterized in that include: Acquire shooting information of the first image and the second image and a first feature matching result; Based on the shooting information and the first feature matching result, a target matrix is ​​calculated to filter out wrong matches in the first feature matching result to obtain a second feature matching result; Connect the associated matching feature points in the second feature matching result to determine whether the slope of each of the connecting lines is within a preset range. If not, it is determined that the feature point corresponding to the connecting line is incorrectly matched.

2. According to claim 1, a feature matching error verification method based on feature point connection, It is characterized in that Acquiring shooting information of the first image and the second image and a first feature matching result, specifically including: receiving the first image and the second image, and extracting matching feature points between the first image and the second image, recording the matching feature points as a first feature matching result; Acquire shooting information of the first image and the second image, the shooting information including: a planar image and a non-planar image; in the planar image, the feature points are in the same plane; in the non-planar image, the feature points are not in the same plane.

3. According to claim 2, a feature matching error verification method based on feature point connection, It is characterized in that Based on the shooting information and the first feature matching result, a target matrix is ​​calculated, specifically including: If the first image and the second image are the planar images, a first target matrix is ​​calculated based on the first feature matching result; If the first image and the second image are the non-planar images, a second target matrix is ​​calculated based on the first feature matching result.

4. According to claim 3, a feature matching error verification method based on feature point connection, It is characterized in that Eliminating the wrong matches in the first feature matching result to obtain the second feature matching result specifically includes: When the first image and the second image are the planar images, verifying each matched feature point in the first feature matching result by using the first target matrix; Each matching feature point that has not passed the verification of the first target matrix is ​​screened out to obtain the second feature matching result.

5. According to claim 3, a feature matching error verification method based on feature point connection, It is characterized in that Screening out the wrong matches in the first feature matching result to obtain the second feature matching result, specifically also includes: When the first image and the second image are non-planar images, verifying each matched feature point in the first feature matching result by using the second target matrix; Each matching feature point that has not passed the verification of the second target matrix is ​​screened out to obtain the second feature matching result.

6. A method for verifying feature matching errors based on feature point connection lines according to claim 4 or 5, It is characterized in that Determining whether the slope of each of the connecting lines exceeds a preset range specifically includes: Calculating the slope corresponding to the line connecting each associated matching point in the second feature matching result and the average of the first slopes corresponding to all matching points; Filtering a target slope value set whose difference between the slope of all matching point connecting lines and the first slope mean is within a first preset slope range; Based on the target slope set, a second slope mean and a standard deviation are calculated, and the preset range is established with the standard deviation to confirm whether the slope of each of the connecting lines exceeds the preset range.

7. The method for verifying feature matching errors based on feature point connection lines according to claim 6, It is characterized in that Based on the target slope set, a second slope mean and a standard deviation are calculated, and the preset range is established with the standard deviation to confirm whether the slope of each of the connecting lines exceeds the preset range, specifically including: Determine whether the difference between the slope corresponding to the connecting line of each associated matching point in the second feature matching result and the second slope average is within the preset range, and if not, confirm that the matching of the feature point corresponding to the connecting line is wrong; If yes, it is confirmed that the feature points corresponding to the connecting line are matched correctly.

8. The method for verifying feature matching errors based on feature point connection lines according to claim 4, It is characterized in that The first target matrix is ​​a homography matrix calculated based on the matching feature points in the first feature matching result.

9. The method for verifying feature matching errors based on feature point connection lines according to claim 5, It is characterized in that The second target matrix is ​​a basic matrix calculated based on the matching feature points in the first feature matching result.