A fast identification method for cosmic muon positioning feature patterns

Abstract

The present invention discloses a method for rapidly identifying a cosmic muon positioning feature map, comprising the following steps: S1, continuously collecting muon scattering data on each navigation segment to obtain a cosmic muon imaging feature image and construct a navigation path imaging data set; S2, extracting a Harris corner point list and a binary descriptor thereof for a pair of cosmic muon imaging feature images at the same position but different times; S3, performing feature point matching on the Harris corner points of two successive cosmic muon imaging feature images at the same position but different times; S4, visualizing the image according to the feature matching result to obtain a cosmic muon imaging feature image of an imaging object with a clear boundary; the method achieves a substantial improvement in the recognition speed of cosmic muon imaging, ensuring real-time navigation under the requirements of cosmic muon imaging; and verified by practical application, the method can achieve a feature recognition effect equivalent to that of a muon positioning feature map ten times the time in a shorter time.

Description

Technical Field

[0001] The present invention relates to the technical fields of navigation, guidance and control, and image recognition and processing, and in particular to a method for quickly identifying cosmic muon positioning feature maps. Background Art

[0002] Navigation and guidance technology is a core area of ​​modern positioning, imaging, and guidance systems, with particular applications in path navigation and rapid imaging in complex environments. Traditional navigation and guidance methods are often limited in environments with significant obstruction or confined spaces, making efficient path planning and positioning imaging difficult. Furthermore, the complex media within confined spaces can lead to significant signal attenuation and reflection interference, making the need for rapid and accurate imaging in such environments even more pressing.

[0003] Cosmic muons are naturally occurring particles with exceptionally strong penetrating power, capable of traversing complex media layers within confined spaces. This property makes cosmic muon-based positioning signatures a novel imaging solution, enabling navigation and positioning in confined spaces and complex media scenarios, with significant value in navigation and guidance. Current research demonstrates that detecting cosmic muon trajectories can reconstruct the spatial structure of confined spaces, providing an effective approach for space imaging.

[0004] However, most existing muon imaging solutions are based on long-term cumulative detection and data processing, which is time-consuming and slow to respond, making it difficult to meet the needs of real-time navigation and guidance in complex scenarios. For example, traditional muon imaging technology usually relies on the collection and processing of large amounts of data to generate clear imaging results. This approach has a sluggish response for application scenarios with high real-time requirements and is not suitable for efficient and fast guidance tasks. Therefore, existing technologies have obvious shortcomings in real-time responsiveness and imaging efficiency, which limits their application potential for navigation and rapid imaging in complex and confined spaces. Based on this, it is necessary to develop a muon imaging technology that can significantly improve imaging speed and responsiveness to meet the navigation and guidance needs in complex media scenarios such as confined spaces. Summary of the Invention

[0005] The purpose of the present invention is to provide a method for rapidly identifying cosmic muon positioning characteristic patterns that solves the above-mentioned technical problems and significantly improves imaging speed and response capability.

[0006] To this end, the technical solution of the present invention is as follows:

[0007] A method for rapidly identifying cosmic muon positioning characteristic patterns, comprising the following steps:

[0008] S1. Divide the navigation path into multiple navigation segments and continuously collect muon scattering data on each navigation segment, including the time when the muon enters and exits the navigation segment, as well as the position of the muon entering and exiting the navigation segment; sequentially use the accumulated muon scattering data for a time length T in each navigation segment for imaging, where T is set to 20 minutes to 100 minutes; name the cosmic muon imaging feature map obtained by imaging "imaging time a-position b" and use it to construct the navigation path imaging dataset;

[0009] S2. Processing a pair of cosmic muon imaging feature images formed successively at the same position and different times obtained in step S1 to extract a list of Harris corner points and a binary descriptor of each Harris corner point in the cosmic muon imaging feature image;

[0010] S3, extracting Harris corner points from the two cosmic muon imaging feature images in step S2, performing feature point matching, and obtaining a list of multiple feature matching point pairs between the two images;

[0011] S4. Visualize the cosmic muon imaging feature image according to the matching result obtained in step S3 to obtain a cosmic muon imaging feature image of the imaging object with a clear boundary.

[0012] Furthermore, in step S1, the method for collecting muon scattering data is: laying upper and lower liquid scintillator muon detectors in each navigation segment; in each navigation segment, the liquid scintillator muon detectors located in the upper and lower paths are arranged in a linear and equidistant manner, and the scattering data collection areas of two adjacent liquid scintillator muon detectors have spatial overlap.

[0013] Furthermore, in step S1, the steps of imaging the muon scattering data accumulated for a time length T in any navigation segment are as follows:

[0014] 1) According to the position of each muon entering and exiting the navigation segment, the scattering displacement of each muon passing through the navigation segment is obtained and scattering angle ;

[0015] 2) Based on the maximum likelihood sequential detection inversion and reconstruction algorithm, the scattering angle and scattering displacement distribution of muons are statistically analyzed, and a likelihood function is defined to solve the scattering density;

[0016] 3) Divide the navigation segment into a voxel grid and, based on the likelihood function defined in step 2), solve for the scattering density of each voxel within each navigation segment;

[0017] 4) Project the scattering density of each voxel obtained in step 3) into a two-dimensional plane scattering density distribution, and map the scattering density into color values ​​through color mapping settings to generate a cosmic muon imaging feature map.

[0018] Furthermore, in step S1, the scattering density in any navigation segment is solved as follows:

[0019] 1) Define the likelihood function as: given the scattering density When all scattering information The probability of is expressed as:

[0020] ,

[0021] Where, is the covariance relationship between the scattering angle and the scattering displacement, which is used to reflect the scattering characteristics of the object to the muon; To measure the extent to which the scattering data deviate from the model prediction; is the scattering density;

[0022] 2) Divide the navigation segment into a voxel grid, and assign an initial scattering density value to each voxel;

[0023] 3) Using the maximum expectation algorithm, the scattering density of each voxel is obtained iteratively when the likelihood function reaches its maximum value.

[0024] Furthermore, in step S1, the method for generating the cosmic muon imaging characteristic map is:

[0025] 1) Project the scatter density results of each voxel in the navigation segment into a 2D scatter density distribution result, and determine the maximum and minimum scatter density values;

[0026] 2) Divide the scattering density range from the minimum scattering density value to the maximum scattering density value into multiple scattering density intervals, and set a color tone and a corresponding color mapping value range for each scattering density interval;

[0027] 3) Mapping the scattering density values ​​of each scattering density interval to the color mapping value range of the corresponding hue;

[0028] 4) Map the scattering density distribution of the two-dimensional plane in step 1) into the color distribution of the two-dimensional plane to generate a two-dimensional cosmic muon imaging feature map.

[0029] Furthermore, the specific implementation steps of step S2 are:

[0030] S201. Using the FAST algorithm, obtain candidate corner points in the cosmic muon imaging feature map;

[0031] S202, performing Harris corner point determination on each candidate corner point;

[0032] S203, using a non-maximum suppression method to remove redundant Harris corner points, and the remaining Harris corner points are significant Harris corner points;

[0033] S204, calculating the main direction of each significant Harris corner point;

[0034] S205. Set a neighbor corner point selection rule based on the main direction of the significant Harris corner point to select n neighbor corner points in the neighborhood of each Harris corner point to form n pairs of comparison pixels; and generate a binary descriptor by comparing the brightness difference between the Harris corner point and the neighbor corner points in the order of selection of the neighbor corner points.

[0035] Furthermore, the specific implementation steps of step S3 are:

[0036] S301, using one cosmic muon imaging feature image as a reference, traverse all Harris corner points in another cosmic muon imaging feature image, and search for the nearest neighbor corner point and the next nearest neighbor corner point for each Harris corner point in the reference image;

[0037] S302: Set a threshold for the nearest neighbor ratio (NNDR) and calculate the nearest neighbor ratio (NNDR) of each Harris corner point in the reference image to screen out reliable matching pairs.

[0038] Furthermore, in step S302, the threshold of NNDR is set to 0.6-0.8, and its calculation formula is:

[0039] ,

[0040] Where, d 1 is the distance between the Harris corner point and its nearest neighbor corner point, d 2 is the distance between the Harris corner point and its next nearest neighbor corner point;

[0041] When the nearest neighbor ratio NNDR of the Harris corner point is less than or equal to the NNDR threshold, the Harris corner point and its nearest neighbor corner point are considered matched. The Harris corner point in the reference image and the nearest neighbor corner point in the other image form a feature matching point pair.

[0042] When the nearest neighbor ratio NNDR of a Harris corner point is greater than the NNDR threshold, the Harris corner point and its nearest neighbor corner point are judged to be mismatched.

[0043] Furthermore, the specific implementation steps of step S4 are:

[0044] S401. Calculate the correspondence matrix between the two images based on the multiple feature matching point pairs obtained in step S3 to align the two cosmic muon imaging feature images to the same coordinate system.

[0045] S402: Based on each pair of feature matching points, draw a straight line between the two aligned images as a matching line, determine the boundary of the matching area based on the area formed by the matching line, and output the boundary position coordinates and the size of the matching area;

[0046] S403. Draw a boundary box on the cosmic muon imaging feature image according to the boundary position coordinates outputted in step S402 to obtain a cosmic muon imaging feature image of the imaging object with a clear boundary.

[0047] Compared with the existing technology, the rapid identification method of the cosmic muon positioning feature map introduces an image matching algorithm to identify the feature points of two short-time coarse imaging cosmic muon imaging feature maps containing spatiotemporal information, ensuring the completeness of the information of the path to be imaged, and then performs feature matching on the feature points between the cosmic muon imaging feature maps to quickly obtain the identification information for reconstructing the imaged object, effectively improving the speed of cosmic muon imaging recognition and ensuring the real-time navigation under the requirements of cosmic muon imaging. After verification in actual application, this method can achieve a feature recognition effect equivalent to ten times the time of the muon positioning feature map in a shorter time. BRIEF DESCRIPTION OF THE DRAWINGS

[0048] Figure 1 This is a flow chart of the method for rapidly identifying cosmic muon positioning characteristic patterns of the present invention;

[0049] Figure 2 This is a schematic diagram of the actual scenario of a certain navigation segment in an embodiment of the present invention;

[0050] Figure 3 Schematic diagram of imaging results obtained by processing cosmic muon scattering data with a collection time of 40 minutes in a certain navigation segment in an embodiment of the present invention through step S1;

[0051] Figure 4 Schematic diagram of imaging results obtained by processing 40 minutes of cosmic muon scattering data collected in a navigation segment in an embodiment of the invention through steps S1 to S4;

[0052] Figure 5 This is a schematic diagram of the imaging results obtained by using a traditional imaging method to collect 400 minutes of cosmic muon scattering data in an actual scenario of a certain navigation segment in an embodiment of the present invention. DETAILED DESCRIPTION

[0053] The present invention will be further described below with reference to the accompanying drawings and specific embodiments, but the following embodiments are by no means intended to limit the present invention in any way.

[0054] See also Figure 1 The specific implementation steps of the method for rapid identification of cosmic muon positioning feature maps are described as follows.

[0055] S1. Construct a navigation path imaging dataset.

[0056] The specific implementation steps of step S1 are described as follows.

[0057] S101. Divide the navigation path into multiple navigation segments, and lay two upper and lower liquid scintillator muon detectors in each navigation segment to continuously collect muon scattering data on each navigation segment.

[0058] In step S101, in order to meet the needs of navigation and positioning in a confined space, several liquid scintillator muon detectors are laid out on the upper and lower sides of each navigation segment along the extension direction of the navigation path to obtain the scattering data of muons before and after passing through each navigation segment.

[0059] In order to ensure the continuity of the scattering imaging data in the imaging results, for each navigation segment, the liquid scintillator muon detectors located on the upper and lower roads are arranged in a linear and equidistant manner, and the spacing between two adjacent liquid scintillator muon detectors on the same road is determined according to the detection range of the liquid scintillator muon detector to ensure that the scattering data collection area has a certain spatial overlap.

[0060] like Figure 2 The figure shows a schematic diagram of the actual scene setting of a navigation segment 1; wherein, two cubic blocks 2 connected at their vertex angles are arranged in the closed space of the navigation segment 1, and three liquid scintillator muon detectors 3 are arranged at equal intervals on the upper and lower sides of the closed space to obtain muon scattering data.

[0061] Since muon scattering imaging is achieved by accumulating muon scattering data, if muon scattering imaging is used to achieve clear imaging of objects in a closed area, it is necessary to accumulate muon scattering data for a sufficient period of time; Figure 5 The traditional method is shown as follows Figure 2 The cosmic muon imaging performed in the scenario shown requires processing of muon scattering data accumulated over a period of 400 minutes to form an effective cosmic muon image. The image can clearly show the shapes of the two cubic blocks 2 within the closed navigation segment 1, with clear boundaries, reflecting imaging details that can only be obtained with data over a longer period of time. In other words, in traditional muon scattering imaging technology, accurate muon image identification information can only be obtained through time.

[0062] The method of the present application aims to realize the rapid identification of the object to be imaged in the closed navigation area, that is, to realize the rapid identification of the object to be imaged by using the muon scattering data accumulated in a short time. Figure 5 Based on this, in this step, the time duration T of the muon scattering data is generally set to 20 minutes to 100 minutes, specifically adjusted according to the muon flux within the navigation area. When the muon flux in the navigation area is high, the acquisition time can be appropriately shortened; conversely, it can be appropriately extended to balance the requirements of data volume and imaging resolution. In this embodiment, the time duration T is set to 40 minutes. That is, during the process of continuously acquiring muon scattering data on each navigation segment, the muon scattering data for each navigation segment is acquired with a 40-minute acquisition cycle and used for muon scattering imaging.

[0063] In this step, the muon scattering data that needs to be acquired for each navigation segment specifically include: ① The time when each muon enters and exits the navigation segment t i and t i ’ ,②The position of each muon entering and exiting the navigation segment ( x i , y i )and( x i ’ , y i ’ ); i is the number of the muon; data ① is used to identify the acquisition time of the muon scattering data to determine whether it is the muon scattering data of the same time length T; data ② is used to identify the navigation segment corresponding to the muon scattering data, and to determine the position of each muon entering and exiting the navigation segment according to the position of each muon entering and exiting the navigation segment ( x i , y i )and( x i ’ , y i ’ ), and the spatial displacement caused by scattering of each muon during its passage through the navigation segment is obtained, namely, the scattering displacement , and the angle change caused by scattering of each muon passing through the navigation segment, that is, the scattering angle .

[0064] S102. Based on the maximum likelihood sequence detection inversion and reconstruction algorithm, the scattering angle and scattering displacement distribution of the muons are statistically analyzed, and a likelihood function for solving the scattering density is defined.

[0065] The specific operation steps of step S102 are as follows:

[0066] Based on the characteristics that muon scattering is affected by statistical factors (such as object material, density distribution, etc.), the scattering displacement and scattering angle of muons conform to the multivariate Gaussian distribution; therefore, the scattering information of any muon passing through the navigation area is The expression is:

[0067] ,

[0068] Where, is the scattering angle, is the scattering displacement;

[0069] Then, the likelihood function is defined, which is used to express the probability of a given scattering density When all scattering information The probability of ; where the likelihood function The expression is:

[0070] ,

[0071] Where, is the covariance relationship between the scattering angle and the scattering displacement, which is used to reflect the scattering characteristics of the object to the muon; To measure the extent to which the scattering data deviate from the model prediction; is the scattering density, which is the variable to be solved by the likelihood function.

[0072] S103 : Based on the likelihood function defined in step S102 , the scattering density of each voxel in each navigation segment is obtained.

[0073] For any navigation segment, the specific processing steps of step S103 are:

[0074] 1) Divide the navigation segment into a voxel grid, and assign an initial scatter density value to each voxel. The accuracy of the voxel grid division is determined according to actual needs to meet the resolution requirements of two-dimensional imaging. In this embodiment, the scatter density of each voxel is initially assigned to 0.

[0075] 2) Based on the likelihood function defined in step S102, an expectation maximum (EM) algorithm is used to iteratively solve the scattering density of each voxel when the likelihood function reaches a maximum value;

[0076] S104, projecting the scattering density of each voxel in each navigation segment obtained in step S103 into a two-dimensional plane scattering density distribution, mapping the scattering density into a color value through color mapping settings, generating and saving a cosmic muon imaging feature map;

[0077] For any navigation segment, the specific processing steps of step S104 are:

[0078] 1) Projecting the scatter density results of each voxel in the navigation segment obtained in step S103 into a scatter density distribution result on a two-dimensional plane, and finding the maximum and minimum values ​​of the scatter density. In this embodiment, the two-dimensional plane is specifically a horizontal plane, and the scatter density distribution result of each pixel on the two-dimensional plane is the cumulative scatter density result of the longitudinal voxels.

[0079] 2) Dividing the scattering density region from the minimum scattering density to the maximum scattering density into three equal intervals, and dividing the scattering density region from the minimum scattering density to the maximum scattering density into a first scattering density region, a second scattering density region, and a third scattering density region in sequence; assigning a color tone and a corresponding color mapping value range to each scattering density region so that the difference in scattering density can be clearly identified after imaging;

[0080] In this embodiment, the first scattering density interval is a low scattering density interval, which corresponds to a warm tone, specifically red, and its GRB range is (200, 0, 0) to (255, 50, 50); the second scattering density interval is a medium scattering density interval, which corresponds to a neutral tone, specifically green, and its GRB range is (0, 200, 0) to (50, 255, 50); the third scattering density interval is a high scattering density interval, which corresponds to a cool tone, specifically blue, and its GRB range is (0, 0, 200) to (50, 50, 255);

[0081] 3) Mapping the scattering density values ​​of each scattering density interval to the color mapping value range of the corresponding hue;

[0082] In this embodiment, the mapping method specifically adopts a linear interpolation algorithm, and its calculation formula is:

[0083] C j = C s +(( ρ i – ρ min ) / ( ρ max – ρ min ))×( C e - C s ),

[0084] Where, C j For the imaging object j The color value of each pixel, Cs is the starting value of the color interval, C e is the color end value of the color interval, ρ min is the minimum scattering density in the scattering density interval, ρ max is the minimum scattering density in the scattering density interval, ρ i The first j Pixel scattering density value;

[0085] 4) Mapping the density distribution of the two-dimensional plane obtained in step 1) above into a color distribution of the two-dimensional plane by replacing the density value with the color value, and generating a two-dimensional cosmic muon imaging feature map based on the two-dimensional plane coordinate system constructed by the navigation segment on the horizontal plane. In this embodiment, the two-dimensional plane coordinate system can be constructed with the center point of the navigation segment as the origin, the extension direction of the navigation path as the x-axis, and the width direction of the navigation path as the y-axis. The pixel size of the image is the projection size of the voxel on the horizontal plane.

[0086] 5) Save the cosmic muon imaging feature map obtained in step 4) with the image name "imaging time a - position b" and add it to the navigation path imaging image dataset; where imaging time a is the start time of muon scattering data acquisition, and position b is the position name or center position coordinate that can be used to identify the navigation segment.

[0087] In this embodiment, step S1 of the present invention is used to Figure 2 The scene shown is imaged and the following is obtained: Figure 3 The characteristic map of cosmic muon imaging is shown in Figure 2. Since this characteristic map is formed by processing only the muon scattering data accumulated over a period of 40 minutes, its accuracy depends on the number of muon events collected. The number of cosmic muons in a short collection time is limited, resulting in blurred and unclear object boundaries. Therefore, in Figure 3 The boundaries and sizes of the two cubic blocks 2 in the navigation segment 1 cannot be clearly identified.

[0088] Through steps S101 through S104, the muon scattering data collected during each navigation segment continuously forms a cosmic muon imaging feature map. These cosmic muon imaging feature maps, taken at different times during each navigation segment, serve as navigation positioning data, forming a navigation path imaging dataset comprised of these cosmic muon imaging feature maps. Furthermore, during subsequent navigation path planning, the cosmic muon imaging feature images acquired at different times at a specific location can be quickly generated based on the imaging requirements of the specified navigation segment.

[0089] S2. Process the cosmic muon imaging feature image obtained in step S1 at the same position and different times to extract the Harris corner point list and the rotation-invariant binary descriptor of each Harris corner point in the cosmic muon imaging feature image.

[0090] As mentioned above, the number of cosmic muons within a short acquisition time is limited, making it impossible to achieve clear imaging of the object. Therefore, in order to obtain an imaging feature map with higher recognition efficiency within the limited imaging time T, the cosmic muon imaging feature map obtained in step S1 needs to be further processed. Specifically, in this step, two cosmic muon imaging feature images formed successively need to be selected and processed separately.

[0091] In step S2, the specific implementation steps of generating a binary descriptor based on any cosmic muon imaging feature map are described as follows.

[0092] S201. Use the FAST algorithm to quickly compare the brightness difference between the central pixel and its surrounding neighboring pixels to obtain candidate corner points in the cosmic muon imaging feature map.

[0093] S202, performing Harris corner point determination on each candidate corner point;

[0094] In this embodiment, taking pixel P as a candidate corner point as an example, 16 equally spaced pixels on a circle with a radius of 3 and centered at pixel P are set as the 16 neighboring pixels S of pixel P. i (i=1, 2, …, 16); compare the brightness difference between pixel P and its 16 neighboring pixels; when at least 9 consecutive neighboring pixels meet the following conditions: I(S i )>I(P)+I threshold or I(S i ) <I(P)-I threshold , then the pixel P is determined to be a Harris corner point; in the above two formulas, I(S i ) is the pixel brightness of the neighboring corner point, I(P) is the pixel brightness of the candidate corner point, I threshold It is a preset parameter, which is set according to the brightness difference of the image.

[0095] S203, using a non-maximum suppression method to remove redundant corner points from the Harris corner points determined in step S202, and the remaining Harris corner points are salient corner points;

[0096] In this embodiment, a neighborhood S1×S1 of the Harris corner point is set to determine whether each Harris corner point is a local brightness maximum within its neighborhood: if so, the Harris corner point is retained; if not, the Harris corner point is determined to be a redundant corner point and discarded; in this embodiment, the neighborhood S1×S1 is set to a neighborhood 3×3.

[0097] S204: Based on the processing result of step S203, the main directions of the Harris corner points that are the remaining salient corner points are calculated respectively, so as to subsequently generate a rotation-invariant binary descriptor;

[0098] In this embodiment, taking any Harris corner point, pixel P, as an example, it is assumed that any pixel point in the neighborhood S1×S1 of pixel P is i The coordinates of x i , y i ), the brightness value is I ( x i , y i ); wherein the neighborhood S1×S1 is set to be the same as step S203, both of which are neighborhoods 3×3; then the main direction of the Harris corner point P passes through the angle θ It means that its expression is:

[0099] ,

[0100] Where, x p and y p are the pixel horizontal and vertical coordinates of the Harris corner point P, x i and y i are the neighborhood pixels of Harris corner point P i The pixel horizontal coordinate and pixel vertical coordinate, I ( x i , y i ) is the brightness of the neighborhood pixel i.

[0101] S205. Set a neighbor corner point selection rule based on the main direction of the Harris corner point to select n neighbor corner points for each Harris corner point in the neighborhood S2×S2 of each Harris corner point, so that each Harris corner point forms n pairs of comparison pixels with its n neighbor corner points. Generate a binary descriptor by comparing the brightness difference between the Harris corner point and its neighbor corner points according to the selection order of the neighbor corner points:

[0102] ,

[0103] Where, I is the brightness of the pixel, i is the serial number of the Harris corner point, is the Harris corner point, It is the neighbor corner point; For serial number i The Harris corner points correspond to the generated binary descriptors.

[0104] In step S205, the neighborhood S2×S2 is a square area centered on the Harris corner point. The size of S2 is set based on application requirements and is typically a 16×16 pixel area or a 32×32 pixel area. In this embodiment, for each Harris corner point, its neighborhood S2×S2 is set to a 16×16 neighborhood. The n neighboring corner points are selected according to the same rule and are evenly distributed within the neighborhood S2×S2, where n=128.

[0105] After the above steps S201 to S205, the two cosmic muon imaging feature maps are respectively generated with a Harris corner point list and a rotation-invariant binary descriptor of each Harris corner point.

[0106] S3. Based on step S2, the feature images of cosmic muon imaging at the same position (i.e., the same navigation segment) at different times are extracted to obtain a Harris corner point list and a binary descriptor of each Harris corner point, and feature point matching is performed to obtain a list of multiple feature matching point pairs between the two images.

[0107] The specific implementation steps of step S3 are described as follows.

[0108] S301. Using one cosmic muon imaging feature image as a reference image, traverse all Harris corner points in another cosmic muon imaging feature image to search for the nearest neighbor corner point and the next nearest neighbor corner point for each Harris corner point in the reference image through nearest neighbor search. The nearest neighbor corner point and the next nearest neighbor corner point specifically refer to: using the binary descriptor of any specified Harris corner point as a reference, the two corner points in the other image with the shortest and second shortest distances from the binary descriptor to the Harris corner point, and the above distance specifically refers to the Hamming distance between the two binary descriptors.

[0109] In this step, the nearest neighbor search can be implemented using the FLANN (Fast Library for Approximate Nearest Neighbors) algorithm by constructing a tree structure of feature descriptors to accelerate the nearest neighbor search. Specifically, the tree structure can adopt a KD tree (k-dimensional tree) or a randomized kd-tree to accelerate the nearest neighbor search in high-dimensional space, so that the nearest neighbor corner points and the next nearest neighbor corner points of a specified Harris corner point in a large number of feature descriptors can be quickly found.

[0110] S302, set the threshold of the nearest neighbor ratio NNDR, and calculate the nearest neighbor ratio NNDR of each Harris corner point in the reference image to screen out reliable matching pairs; specifically,

[0111] Calculate the nearest neighbor ratio NNDR of Harris corner points, and the calculation formula is:

[0112] ,

[0113] Where, d 1 is the distance between the Harris corner point and its nearest neighbor corner point, d 2 is the distance between the Harris corner point and its next nearest neighbor corner point;

[0114] In this embodiment, the threshold of the nearest neighbor ratio NNDR is set to 0.75, and the following determination is made:

[0115] ① When the nearest neighbor ratio (NNDR) of a Harris corner point is less than or equal to 0.75, the Harris corner point and its nearest neighbor corner point are considered matched. Furthermore, the Harris corner point in the reference image and the nearest neighbor corner point in the other image form a feature matching point pair.

[0116] ② When the nearest neighbor ratio (NNDR) of a Harris corner point is greater than 0.75, the Harris corner point and its nearest neighbor corner point are considered mismatched.

[0117] After the above steps S301 and S302, multiple feature matching point pairs are screened out from the two cosmic muon imaging feature images, that is, a list of multiple feature matching point pairs between each pair of images.

[0118] S4. Perform visualization processing based on the matching result obtained in step S3 to obtain a cosmic muon imaging characteristic image of the imaging object with a clear boundary.

[0119] The specific implementation steps of step S4 are described as follows.

[0120] S401. Calculate the correspondence matrix between the two images based on the multiple feature matching point pairs obtained in step S3 to align the two cosmic muon imaging feature images to the same coordinate system.

[0121] Specifically, this step can be implemented by the RANSAC algorithm, and the robustness of the matrix calculation is ensured by the RANSAC algorithm; wherein the matrix adopts the homography matrix, and its calculation expression is:

[0122] ,

[0123] Where δ is the discriminant function, specifically the distance function; p i and pi′ are the two corresponding feature matching corner points from the reference image and the other image respectively;

[0124] Then, based on the homography matrix H obtained above, all pixels in the reference image are mapped to the coordinate system of the other image through matrix multiplication and homogeneous coordinate transformation to achieve image alignment, which is expressed as:

[0125] ,

[0126] Where x, y are the pixel coordinates of the feature matching corner point in the reference image, and x′, y′ are the pixel coordinates of the feature matching corner point in the reference image in the other image.

[0127] As a preferred technical solution of this embodiment, a sharpening filter (such as a Laplacian operator) is used to process the aligned images to enhance the details of the common area.

[0128] S402: Based on each pair of feature matching points, a straight line is drawn between the two aligned images as a matching line to make the matching relationship between the two images intuitively visible; then, based on the area formed by the matching line, the boundary of the matching area is determined, and the boundary position coordinates and the size of the matching area are output;

[0129] In this step, a sharpening filter may be further used to process the bounding box information to enhance edges and details in the image and improve the clarity of the matching area.

[0130] S403. Draw the bounding box position coordinates output by the above step S402 on the cosmic muon imaging feature image to obtain a cosmic muon imaging feature image with a clear boundary imaging object.

[0131] like Figure 4 The figure shows the imaging characteristics of cosmic muons for 40 minutes obtained by the method of the present invention based on the coarse image processing of cosmic muons. Figure 5Compared with the processing results, the present application obtains a cosmic muon coarse image based on muon scattering data with a time length of 40 minutes. After the image feature recognition, matching and visualization processing of the present invention, the coarse image of the cosmic muon can be obtained. Figure 5 Schematic diagram of cosmic muon imaging features with the same imaging effect obtained using 400 minutes of scattering data.

[0132] In summary, compared with the traditional method of obtaining accurate muon image recognition information only through time, the method of the present application can achieve data effects close to the long-term imaging of the traditional method in a shorter time; specifically, see Figure 4 and Figure 5 The processing effect of the method of the present invention is that a faster muon positioning feature map acquisition effect is achieved, which is equivalent to shortening the imaging time to one tenth, proving that the method of the present invention has significant advantages in feature recognition efficiency and accuracy, that is, the same pattern recognition effect as 10 times the imaging time can be achieved in a shorter time.

Claims

1. A method for rapidly identifying cosmic muon positioning characteristic patterns, characterized in that: The steps are: S1. Divide the navigation path into multiple navigation segments and continuously collect muon scattering data on each navigation segment, including the time when the muon enters and exits the navigation segment, as well as the position of the muon entering and exiting the navigation segment. The muon scattering data accumulated for a time length T in each navigation segment are sequentially used for imaging, where T is set to 20 minutes to 100 minutes. The imaging feature map of the cosmic muon obtained by imaging is named "imaging time a - position b" and used to construct the navigation path imaging dataset. S2. Processing a pair of cosmic muon imaging feature images formed successively at the same position and different times obtained in step S1 to extract a list of Harris corner points and a binary descriptor of each Harris corner point in the cosmic muon imaging feature image; S3, extracting Harris corner points from the two cosmic muon imaging feature images in step S2, performing feature point matching, and obtaining a list of multiple feature matching point pairs between the two images; S4. Visualize the cosmic muon imaging feature image according to the matching result obtained in step S3 to obtain a cosmic muon imaging feature image of the imaging object with a clear boundary.

2. The method for rapid identification of cosmic muon positioning characteristic patterns according to claim 1, characterized in that: In step S1, the method for collecting muon scattering data is as follows: liquid scintillator muon detectors are laid out in upper and lower paths in each navigation segment; in each navigation segment, the liquid scintillator muon detectors located in the upper and lower paths are arranged in a linear and equidistant manner, and the scattering data collection areas of two adjacent liquid scintillator muon detectors have spatial overlap.

3. The method for rapid identification of cosmic muon positioning characteristic patterns according to claim 1, characterized in that: In step S1, the steps of imaging the muon scattering data accumulated for a time length T in any navigation segment are as follows: 1) According to the position of each muon entering and exiting the navigation segment, the scattering displacement of each muon passing through the navigation segment is obtained and scattering angle ; 2) Based on the maximum likelihood sequential detection inversion and reconstruction algorithm, the scattering angle and scattering displacement distribution of muons are statistically analyzed, and a likelihood function is defined to solve the scattering density; 3) Divide the navigation segment into a voxel grid and, based on the likelihood function defined in step 2), solve for the scattering density of each voxel within each navigation segment; 4) Project the scattering density of each voxel obtained in step 3) into a two-dimensional plane scattering density distribution, and map the scattering density into color values ​​through color mapping settings to generate a cosmic muon imaging feature map.

4. The method for rapid identification of cosmic muon positioning characteristic patterns according to claim 3, characterized in that: In step S1, the scattering density solution in any navigation segment is: 1) Define the likelihood function as: given the scattering density When all scattering information The probability of is expressed as: ， Where, is the covariance relationship between the scattering angle and the scattering displacement, which is used to reflect the scattering characteristics of the object to the muon; To measure the extent to which the scattering data deviate from the model prediction; is the scattering density; 2) Divide the navigation segment into a voxel grid, and assign an initial scattering density value to each voxel; 3) Using the maximum expectation algorithm, the scattering density of each voxel is obtained iteratively when the likelihood function reaches its maximum value.

5. The method for rapid identification of cosmic muon positioning characteristic patterns according to claim 3, characterized in that: In step S1, the method for generating the cosmic muon imaging feature map is: 1) Project the scatter density results of each voxel in the navigation segment into a 2D scatter density distribution result, and determine the maximum and minimum scatter density values; 2) Divide the scattering density range from the minimum scattering density value to the maximum scattering density value into multiple scattering density intervals, and set a color tone and a corresponding color mapping value range for each scattering density interval; 3) Mapping the scattering density values ​​of each scattering density interval to the color mapping value range of the corresponding hue; 4) Map the scattering density distribution of the two-dimensional plane in step 1) into the color distribution of the two-dimensional plane to generate a two-dimensional cosmic muon imaging feature map.

6. The method for rapid identification of cosmic muon positioning characteristic patterns according to claim 1, characterized in that: The specific implementation steps of step S2 are: S201. Using the FAST algorithm, obtain candidate corner points in the cosmic muon imaging feature map; S202, performing Harris corner point determination on each candidate corner point; S203, using a non-maximum suppression method to remove redundant Harris corner points, and the remaining Harris corner points are significant Harris corner points; S204, calculating the main direction of each significant Harris corner point; S205. Set a neighbor corner point selection rule based on the main direction of the significant Harris corner point to select n neighbor corner points in the neighborhood of each Harris corner point to form n pairs of comparison pixels; and generate a binary descriptor by comparing the brightness difference between the Harris corner point and the neighbor corner points in the order of selection of the neighbor corner points.

7. The method for rapid identification of cosmic muon positioning characteristic patterns according to claim 6, characterized in that: The specific implementation steps of step S3 are: S301, using one cosmic muon imaging feature image as a reference, traverse all Harris corner points in another cosmic muon imaging feature image, and search for the nearest neighbor corner point and the next nearest neighbor corner point for each Harris corner point in the reference image; S302: Set a threshold for the nearest neighbor ratio (NNDR) and calculate the nearest neighbor ratio (NNDR) of each Harris corner point in the reference image to screen out reliable matching pairs.

8. The method for rapid identification of cosmic muon positioning characteristic patterns according to claim 7, characterized in that: In step S302, the threshold of NNDR is set to 0.6-0.8, and its calculation formula is: ， Where, d 1 is the distance between the Harris corner point and its nearest neighbor corner point, d 2 is the distance between the Harris corner point and its next nearest neighbor corner point; When the nearest neighbor ratio NNDR of the Harris corner point is less than or equal to the NNDR threshold, the Harris corner point and its nearest neighbor corner point are considered matched. The Harris corner point in the reference image and the nearest neighbor corner point in the other image form a feature matching point pair. When the nearest neighbor ratio NNDR of a Harris corner point is greater than the NNDR threshold, the Harris corner point and its nearest neighbor corner point are judged to be mismatched.

9. The method for rapid identification of cosmic muon positioning characteristic patterns according to claim 8, characterized in that: The specific implementation steps of step S4 are: S401. Calculate the correspondence matrix between the two images based on the multiple feature matching point pairs obtained in step S3 to align the two cosmic muon imaging feature images to the same coordinate system. S402: Based on each pair of feature matching points, draw a straight line between the two aligned images as a matching line, determine the boundary of the matching area based on the area formed by the matching line, and output the boundary position coordinates and the size of the matching area; S403. Draw a boundary box on the cosmic muon imaging feature image according to the boundary position coordinates outputted in step S402 to obtain a cosmic muon imaging feature image of the imaging object with a clear boundary.

Images