Moon polar region sequence image navigation method based on shadow features

Abstract

The invention discloses a moon polar region sequence image navigation method based on shadow features, which comprises the following steps: binarizing two adjacent frames of sequence images, extracting a stable shadow region, performing morphological corrosion processing to obtain each connected domain formed by the shadow region, and calculating the centroid of each connected domain; screening each connected domain by using the area of each connected domain and the distance between the shadow centroid and the image edge to obtain a stable shadow feature, and constructing a descriptor for the shadow feature; matching the shadow features by using the descriptors, judging the matching result, and reserving the centroid meeting the condition as a final matching result; resolving the relative pose of the aircraft between the two adjacent frames of images by using homography matrix decomposition according to shadow feature matching results on the two adjacent images; according to prior height information of the laser sensor, real physical scale information of relative positions among the sequence images is recovered; and repeating the steps on two subsequent adjacent frames in the sequence image to obtain the position information of the aircraft at each imaging point.

Description

Technical Field

[0001] The invention relates to a lunar polar region sequence image navigation method based on shadow features, belonging to the field of visual navigation. Background Art

[0002] At present, visual navigation is generally used to achieve high-precision landing detection of extraterrestrial bodies. Visual navigation positioning methods can be divided into two types according to whether prior data is used. One is a terrain relative navigation method based on prior terrain (or image) data of the planetary surface. This method obtains information such as the position and attitude of the aircraft by matching features between the known planetary surface terrain map and the image taken by the probe navigation camera; the other method is to match the adjacent frames of the sequence navigation images taken by the aircraft in the absence of prior terrain data to obtain the relative position, attitude and other information of the aircraft. Due to the limitations of factors such as the accuracy and resolution of prior terrain data, the application scenarios of the sequence image frame navigation method that does not rely on prior data are wider. In the process of navigation between sequence image frames, it is necessary to extract matching features between adjacent frame images. The following problems exist in this process:

[0003] (1) The low solar altitude angle in the lunar polar region results in many shadows in the captured images, which leads to insufficient stability and matching robustness in feature extraction from sequence images;

[0004] (2) The algorithm takes a long time and is not conducive to real-time navigation. Summary of the invention

[0005] The technical problem solved by the present invention is: to overcome the shortcomings of the prior art and provide a lunar polar region sequence image navigation method based on shadow features, which can improve the stability of feature extraction and the robustness of matching, while improving the calculation efficiency and meeting the needs of real-time navigation.

[0006] The technical solution of the present invention is: a lunar polar region sequence image navigation method based on shadow features, comprising:

[0007] S1 binarizes the two adjacent frames of the sequence image when the navigation starts, and extracts the shadow area of ​​the image;

[0008] S2 performs morphological corrosion processing on the obtained shadow area, and obtains the centroid of each connected domain formed by the shadow area after the corrosion processing;

[0009] S3 uses the area size of each connected domain formed by the shadow area and the distance between the shadow centroid and the image edge as screening conditions to screen each connected domain and obtain a stable shadow feature;

[0010] S4 constructs a shadow feature descriptor for each stable shadow feature to obtain a shadow feature description matrix of each shadow feature;

[0011] S5 uses the shadow feature description matrix to match the shadow features of two adjacent frame sequence images, and further screens the matching results, retaining the shadow feature matching results that meet the conditions as the final matching results;

[0012] S6 uses the final matching results and the homography matrix decomposition method to solve the relative position of the aircraft between two adjacent frames of images; and restores the real physical scale information of the relative position between the sequence images based on the prior height information of the laser sensor;

[0013] S7 repeats the process of S1 to S6 for two subsequent adjacent frames in the sequence of images to obtain the position information of the aircraft at each imaging point.

[0014] Preferably, the specific method for binarizing two adjacent frame sequence images is:

[0015] The image is divided into blocks using the grid method, and the Otsu method is used to determine the binarization threshold for the image within each grid range. The area smaller than the threshold is defined as the shadow area, and a binary image is obtained;

[0016] Perform pyramid operation on the binary image to obtain a multi-layer pyramid image;

[0017] Preserve shadows that can be extracted on multiple layers of pyramid images.

[0018] Preferably, a corresponding shadow feature description matrix is ​​constructed for each shadow feature, and the construction method is:

[0019] For a connected domain where a shadow feature A is located, first take its centroid A as the center of the circle, set the centroid of the connected domain where the shadow feature closest to A is located as B, and the distance between the two is r; establish concentric rings with r as the minimum radius and R as the maximum radius, the value of R is an integer multiple of r, and the number of concentric rings is g=R / r;

[0020] Take the vector from the center of mass A to the center of mass B as the +X axis, and the +Y axis as the direction of the +X axis rotated 90° clockwise around the center of mass A. Divide the concentric rings into four quadrants of the coordinate system, and each ring is divided into four sub-rings.

[0021] The situation of shadow features falling into each sub-ring is recorded: when a shadow feature falls into it, the sub-ring is recorded as 1, and when no shadow feature falls into it, the sub-ring is recorded as 0, thus forming a g×4 shadow feature description matrix, where g represents the number of the ring where the sub-ring is located, and 4 represents the four quadrants where the ring is located.

[0022] Preferably, the specific method for matching shadow features is:

[0023] S5-1. For a shadow feature i on the first image, its shadow feature description matrix is ​​subtracted from the shadow feature description matrix of the shadow feature j on the second image, and the position with a score of 0 in the subtracted matrix is ​​scored as G1; the position with a score of -1 or 1 is scored as G2, and finally the total score of all positions in the subtracted matrix is ​​calculated as the matching score;

[0024] S5-2, traverse the shadow feature description matrices of other shadow features in the second image in the manner of S5-1, select and record the shadow features corresponding to the shadow feature description matrices whose total scores are greater than 85%×g×4×G1 as candidate matching features of the current shadow feature;

[0025] S5-3. Repeat the methods of S5-1 to S5-2 to traverse all shadow features on the first image, and obtain all candidate matching features of all shadow features in the first image in the second image.

[0026] Preferably, G1 is a positive integer, G2 is a negative integer, and G1=2×|G2|.

[0027] Preferably, when further screening the matching results, the reprojection error method is first used to perform a preliminary screening of the matching results of all shadow features, and finally RANSAC is used to eliminate the false matches in the shadow feature matching pairs retained after the preliminary screening.

[0028] Preferably, the process of preliminary screening using the reprojection error method is:

[0029] Based on the fact that the distances between all shadow features on the first image and the distances between the corresponding matching features on the second image should be in the same proportion, H pairs of shadow feature matching pairs are randomly selected from all matching features;

[0030] The selected H pairs of shadow feature matching pairs are used to solve the initial homography relationship between the two images, and the initial homography relationship is used to solve the reprojection error of all shadow features. When the reprojection error is less than K pixels, this group of shadow features is considered to be correctly matched and retained; otherwise, it is discarded;

[0031] The values ​​of H and K are determined according to the navigation mission indicators.

[0032] Compared with the prior art, the present invention has the following advantages:

[0033] (1) The present invention extracts shadow areas through the grid pyramid method, effectively improving the stability of shadow feature extraction;

[0034] (2) The present invention effectively improves the accuracy and robustness of feature matching by constructing a shadow feature descriptor with scale and rotation invariance. BRIEF DESCRIPTION OF THE DRAWINGS

[0035] Figure 1 It is a schematic diagram of the algorithm of the present invention;

[0036] Figure 2 Schematic diagram of the shadow feature preprocessing process of the present invention; wherein a is the generated binary image, b is the shadow image after morphological processing, and c is a schematic diagram of the shadow connection domain and centroid extraction results.

[0037] Figure 3 Schematic diagram of the shadow feature screening results of the present invention; a is a schematic diagram of the effect before screening, b is a schematic diagram of the effect after screening,

[0038] Figure 4 A schematic diagram is constructed for the shadow feature descriptor of the present invention;

[0039] Figure 5 Schematic diagram of the matching score judgment process of the present invention. DETAILED DESCRIPTION

[0040] The present invention is further described below in conjunction with the accompanying drawings. In the process of matching between sequence image frames using shadow features, the following steps are included:

[0041] 1. In the process of matching sequence images using shadow features, in order to obtain stable shadow features, a grid pyramid method is used to binarize the navigation image and extract the shadow area of ​​the image. The specific method is:

[0042] First, the image is divided into blocks using the grid method, and the Otsu method is used to determine the binarization threshold for the image within each block. The area less than the threshold is the shadow area. In addition, due to the scale differences between navigation images, in order to ensure that the extracted shadows have a certain stability in different scales, it is necessary to perform pyramid operations on the navigation images to retain the shadows that can be stably extracted at different scales. Each layer of the image is divided into 3×3 grids and binarized using the Otsu method; this paper sets a three-layer pyramid and downsamples the image with a sampling interval of 0.5 times, and only retains the shadow areas on the corresponding original images that can be stably present in the three-layer pyramid images.

[0043] 2. If Figure 2 As shown, the shadow area obtained in 1 is morphologically eroded to delete the smaller shadow area, so as to achieve the purpose of segmenting the unstable shadow. After the erosion process is completed, the area of ​​the connected domain where the remaining shadow area is located and its centroid are obtained to form the shadow feature.

[0044] 3. Use the connected domain area of ​​the shadow feature and the distance between the shadow centroid and the image edge as screening conditions to screen the shadow features. The relevant parameter settings and process are as follows:

[0045] The minimum area of ​​the connected domain is set to 100 pixels and the maximum is set to 20,000 pixels, and the shadow features that are less than 100 pixels away from the edge of the image are deleted. The screening effect is as follows: Figure 3 shown.

[0046] 4. After obtaining stable shadow features, it is necessary to construct a descriptor for each shadow feature, that is, a shadow feature description matrix. The construction process is mainly divided into the following steps:

[0047] (1) For the shadow feature A, first, take the mass center of A as the center of the circle, set the mass center of the shadow feature closest to A as B, and the distance between the two as r. Take r as the minimum radius and R as the maximum radius, whose value is an integer multiple of r, to establish concentric rings. The number of concentric rings is g=R / r; in the embodiment provided by the present invention, R=5r;

[0048] →

[0049] (2) Take BA as the X-axis, +Y-axis is the direction of +X-axis rotating 90° counterclockwise around the center of mass B, and clockwise rotation is the positive direction, such as Figure 4 As shown in FIG. 1 , the concentric rings are divided into four sub-rings at 0° to 90°, 90° to 180°, 180° to 270°, and 270° to 360° according to the four quadrants of the coordinate system. The shadow features falling into each sub-ring are calculated; when a shadow feature falls into the sub-ring, the sub-ring is recorded as 1, and when no shadow feature falls into the sub-ring, the sub-ring is recorded as 0, thereby forming a g×4 shadow feature description matrix, where g represents the number of rings where the sub-ring is located, and 4 represents the four quadrants where the ring is located.

[0050] Finally, the shadow features and their corresponding shadow feature description matrix are obtained.

[0051] 5. After obtaining the shadow features and their feature descriptors on two adjacent frames, set the matching score as the threshold to match the shadow features. Then use RANSAC and reprojection error and other conditions to judge the candidate matching features, and only retain the centroid of the shadow features that meet the conditions as the final matching result.

[0052] The matching method is:

[0053] S5-1, such as Figure 5 As shown, for a shadow feature i on the first image, its shadow feature description matrix is ​​subtracted from the shadow feature description matrix of the shadow feature j on the second image, and the position with a 0 in the subtraction matrix obtained after the subtraction is scored as 10; the position with a -1 or 1 is scored as -5, and finally the total score of all positions is calculated as the matching score;

[0054] S5-2, traversing the shadow feature description matrices of other shadow features in the second image in the manner of S5-1, selecting and recording the shadow features corresponding to the shadow feature description matrices whose total scores are greater than 85%×g×4×10 as candidate matching features for the current shadow feature;

[0055] S5-3. Return to S5-1, and repeat the methods of S5-1 to S5-2 to traverse all shadow features on the first image to obtain all candidate matching features of all shadow features in the first image.

[0056] When judging the matching results, RANSAC and reprojection error are used. The specific method is:

[0057] For all candidate matching features obtained from the two images, the distances between all shadow features on the first image and the distances between the corresponding matching features on the second image should be in the same proportion. Based on this, 6 pairs of shadow feature matching pairs are randomly selected from all matching features.

[0058] The shadow feature matching pairs selected above are used to solve the initial homography relationship between the two images, and the reprojection error of all shadow features is solved using the initial homography relationship. When the reprojection error is less than 5 pixels, this group of shadow features is considered to be correctly matched and retained; otherwise, it is discarded. Repeat the iteration until all shadow feature matching pairs are traversed.

[0059] Finally, RANSAC is used to eliminate the mismatches in the obtained matching point pairs.

[0060] 6. After obtaining the final shadow feature matching results on two adjacent images, the relative position of the aircraft between the two adjacent frames of images is solved by using the homography matrix decomposition; and the real physical scale information of the relative position between the sequence images is restored based on the prior height information of the laser sensor;

[0061] 7. Repeat steps 1-6 for the next two adjacent frames in the sequence of images to obtain the position information of the aircraft at each imaging point.

[0062] The contents not described in detail in the specification of the present invention belong to the prior art known to the professional and technical personnel in this field.

Claims

1. A lunar polar region sequence image navigation method based on shadow features, characterized in that The steps include: S1 binarizes the two adjacent frames of the sequence image when the navigation starts, and extracts the shadow area of ​​the image; S2 performs morphological corrosion processing on the obtained shadow area, and obtains the centroid of each connected domain formed by the shadow area after the corrosion processing; S3 uses the area size of each connected domain formed by the shadow area and the distance between the shadow centroid and the image edge as screening conditions to screen each connected domain and obtain a stable shadow feature; S4 constructs a shadow feature descriptor for each stable shadow feature to obtain a shadow feature description matrix of each shadow feature; S5 uses the shadow feature description matrix to match the shadow features of two adjacent frame sequence images, and further screens the matching results, retaining the shadow feature matching results that meet the conditions as the final matching results; S6 uses the final matching results and the homography matrix decomposition method to solve the relative position of the aircraft between two adjacent frames of images; and restores the real physical scale information of the relative position between the sequence images based on the prior height information of the laser sensor; S7 repeats the process of S1 to S6 for two subsequent adjacent frames in the sequence of images to obtain the position information of the aircraft at each imaging point.

2. The method for lunar polar region sequence image navigation based on shadow features according to claim 1, characterized in that: The specific method for binarizing two adjacent frame sequence images is as follows: The image is divided into blocks using the grid method, and the Otsu method is used to determine the binarization threshold for the image within each grid range. The area smaller than the threshold is defined as the shadow area, and a binary image is obtained; Perform pyramid operation on the binary image to obtain a multi-layer pyramid image; Preserve shadows that can be extracted on multiple layers of pyramid images.

3. The method for lunar polar region sequence image navigation based on shadow features according to claim 1, characterized in that: Construct a corresponding shadow feature description matrix for each shadow feature. The construction method is: For a connected domain where a shadow feature A is located, first take its centroid A as the center of the circle, set the centroid of the connected domain where the shadow feature closest to A is located as B, and the distance between the two is r; establish concentric rings with r as the minimum radius and R as the maximum radius, the value of R is an integer multiple of r, and the number of concentric rings is g=R / r; Take the vector from the center of mass A to the center of mass B as the +X axis, and the +Y axis as the direction of the +X axis rotated 90° clockwise around the center of mass A. Divide the concentric rings into four quadrants of the coordinate system, and each ring is divided into four sub-rings. The situation of shadow features falling into each sub-ring is recorded: when a shadow feature falls into it, the sub-ring is recorded as 1, and when no shadow feature falls into it, the sub-ring is recorded as 0, thus forming a g×4 shadow feature description matrix, where g represents the number of the ring where the sub-ring is located, and 4 represents the four quadrants where the ring is located.

4. The method for lunar polar region sequence image navigation based on shadow features according to claim 3, characterized in that: The specific method for matching shadow features is: S5-1. For a shadow feature i on the first image, its shadow feature description matrix is ​​subtracted from the shadow feature description matrix of the shadow feature j on the second image, and the position with a score of 0 in the subtracted matrix is ​​scored as G1; the position with a score of -1 or 1 is scored as G2, and finally the total score of all positions in the subtracted matrix is ​​calculated as the matching score; S5-2, traverse the shadow feature description matrices of other shadow features in the second image in the manner of S5-1, select and record the shadow features corresponding to the shadow feature description matrices whose total scores are greater than 85%×g×4×G1 as candidate matching features of the current shadow feature; S5-3. Repeat the methods of S5-1 to S5-2 to traverse all shadow features on the first image, and obtain all candidate matching features of all shadow features in the first image in the second image.

5. The method for lunar polar region sequence image navigation based on shadow features according to claim 4, characterized in that: G1 is a positive integer, G2 is a negative integer, G1=2×|G2|.

6. The method for lunar polar region sequence image navigation based on shadow features according to claim 1, characterized in that: When further screening the matching results, the reprojection error method is first used to preliminarily screen the matching results of all shadow features, and finally RANSAC is used to eliminate the false matches in the shadow feature matching pairs retained after the preliminary screening.

7. The method for lunar polar region sequence image navigation based on shadow features according to claim 6, characterized in that: The process of preliminary screening using the reprojection error method is: Based on the fact that the distances between all shadow features on the first image and the distances between the corresponding matching features on the second image should be in the same proportion, H pairs of shadow feature matching pairs are randomly selected from all matching features; The selected H pairs of shadow feature matching pairs are used to solve the initial homography relationship between the two images, and the initial homography relationship is used to solve the reprojection error of all shadow features. When the reprojection error is less than K pixels, this group of shadow features is considered to be correctly matched and retained; otherwise, it is discarded; The values ​​of H and K are determined according to the navigation mission indicators.

Images