Method for rapid detection of position and azimuth angle of multiple moving targets based on single-pixel imaging

Abstract

The application belongs to the field of target tracking, and particularly relates to a multi-motion target position and azimuth fast detection method based on single-pixel imaging, which comprises the following steps: S1, obtaining one-dimensional projection curves of scene images in four directions of 0°, 45°, 90° and 135° for each frame of scene images comprising multiple motion targets; S2, dividing the scene images into N regions by using a one-dimensional projection inverse reconstruction method, and each region has only one motion target; S3, modulating the scene images by using a deformed geometric moment mode to obtain the position coordinates of the center of mass of each motion target; and S4, obtaining a second-order central moment mode according to the center of mass coordinates of each motion target, and modulating the scene images to obtain the azimuth of each motion target. The application can realize fast and high-precision detection of the position coordinates and the azimuth of multiple motion targets.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of moving target tracking, in particular to a method for rapidly detecting the position and azimuth angle of multiple moving targets based on single-pixel imaging. BACKGROUND

[0002] Single-pixel imaging technology is an imaging technology that modulates a target scene and uses a single-pixel detector without spatial resolution to collect the light intensity value reflected or transmitted by the scene. By using the correlation between the modulated speckle and the light intensity value, the image of the target scene can be restored. Compared with the traditional two-dimensional area array imaging method, single-pixel imaging technology has the characteristics of wide spectrum and high sensitivity, and can be applied to wavebands that cannot be responded by area array cameras and low-light imaging.

[0003] There are two main types of methods for tracking moving targets based on single-pixel imaging technology: image-based and imageless tracking methods. The image-based moving target tracking method requires multiple detections to obtain scene images, and then uses image processing algorithms to obtain the position of the moving target. The image-based moving target tracking method has the following disadvantages: first, the imaging time is relatively long, which is not suitable for rapid positioning and tracking of moving targets; second, the positioning accuracy depends on the quality of the reconstructed image, and when the quality of the reconstructed image is poor, the positioning error may be large; in addition, in practical applications, the data storage capacity of the image is large, and it is a challenge for long-time moving target tracking. Therefore, in recent years, more and more attention has been paid to the research of imageless moving target tracking methods based on single-pixel imaging.

[0004] Imageless moving target tracking methods based on single-pixel imaging do not require scene images, and are very suitable for moving target recognition. However, most of the current research is related to the acquisition of motion parameters such as translation, rotation, and translation plus rotation of a single moving target based on single-pixel imaging, and there is still a lack of research on positioning methods for multiple moving targets. Therefore, a method is needed to detect the position and azimuth angle of multiple moving targets based on single-pixel imaging technology. SUMMARY

[0005] In view of the deficiencies of the prior art, the present application proposes a method for detecting the position and azimuth angle of multiple moving targets based on single-pixel imaging to rapidly determine the position and azimuth angle of multiple moving targets, and promotes the development and application of single-pixel imaging technology in the cooperative control of multiple moving targets.

[0006] To achieve the above purpose, the present application discloses a method for rapidly detecting the position and azimuth angle of multiple moving targets based on single-pixel imaging, which comprises the following steps:

[0007] S1: obtaining a one-dimensional projection curve of the scene image for each frame of the scene image including the moving target;

[0008] For each frame of the scene image including the moving target, a first light intensity value is obtained by modulating the scene image using a Fourier mode, and a Fourier coefficient F i (u,v) is obtained according to the first light intensity value using a three-step phase shift formula i (u,v) is the Fourier coefficient of the i-th frame of the scene image at the frequency (u,v); and a one-dimensional projection curve of the scene image in the directions of 0°, 45°, 90° and 135° is obtained according to the Fourier coefficient F i (u,v).

[0009] S2: dividing the scene image into N regions using one-dimensional projection;

[0010] The region where the N moving targets are located is obtained using a projection inverse reconstruction method, and the scene image is regionally segmented according to the position of the moving target region. If there are N moving targets, the scene image is divided into N regions, so that there is only one moving target in each region; wherein N is a positive integer;

[0011] S3: modulating the scene image using a deformed geometric moment mode to obtain the centroid coordinate of each moving target;

[0012] According to the coordinate range of the region where each moving target is located, a corresponding deformed geometric moment mode is generated. After modulating the scene image using the deformed geometric moment mode, a second light intensity value, a third light intensity value and a fourth light intensity value are obtained. According to the second light intensity value, the third light intensity value and the fourth light intensity value, the centroid coordinate of the moving target relative to the region where it is located is obtained, and then converted into the centroid coordinate of the moving target in the scene image;

[0013] S4: obtaining a corresponding azimuth angle according to the centroid coordinate of each moving target;

[0014] According to the centroid coordinate of the moving target in S3, a second-order central moment mode is obtained. The normalized difference method is used on the second-order central moment mode to obtain a normalized difference second-order central moment mode. The scene image is modulated using the normalized difference second-order central moment mode to obtain a fifth light intensity value and a sixth light intensity value. According to the fifth light intensity value and the sixth light intensity value, a second-order central moment is obtained. According to the second-order central moment, a covariance matrix is constructed. Two eigenvalues and two corresponding eigenvectors of the covariance matrix are obtained. According to the eigenvectors, the azimuth angle of the moving target is obtained.

[0015] Preferably, in step S1, for each frame of the scene image including the moving target, the one-dimensional projection curve of the scene image is obtained as follows:

[0016] S11: generating a series of different spatial frequencies and initial phases respectively gray-scale Fourier mode

[0017]

[0018] wherein a is the average light intensity value of the gray-scale Fourier mode, b is the contrast of the gray-scale Fourier mode, x, y are the spatial coordinates of the scene image respectively, u, v are the spatial frequencies of the discrete domain respectively; when the size of the scene image of the moving target is M x N, the values of x are 1, 2, …, N respectively, the values of y are 1, 2, …, M respectively, the values of u are the values of v are is the initial phase, the values are

[0019] S12: modulating the scene image of the moving target using the binary Fourier mode to obtain a first modulated image;

[0020] S13: collecting a first light intensity value of the first modulated image using a single-pixel detector;

[0021]

[0022] wherein e is the background noise, I i (x, y) is the i-th scene image; is the binary Fourier mode, is the first light intensity value;

[0023] S14: calculating the Fourier coefficient F i (u, v) at the frequency (u, v) using the three-step phase shift method formula:

[0024]

[0025] S15: obtaining the one-dimensional projection curve of the scene image in the 0°, 45°, 90°, 135° directions;

[0026] The one-dimensional projection of the scene in the 0°, 45°, 90°, 135° directions is equal to the one-dimensional inverse Fourier transform result of the corresponding 0°, 45°, 90°, 135° direction slice of the two-dimensional Fourier transform of the image over the origin:

[0027]

[0028] wherein θ is the included angle between the projection direction and the x-axis direction, T i (θ) is the projection curve of the i-th scene image in the θ direction, F i -1 is the Fourier coefficient Fi a one-dimensional inverse Fourier transform of (u, v).

[0029] Preferably, the scene image is divided into N regions by one-dimensional projection in step S2, specifically:

[0030] S21: preliminary obtain the regions where the multiple moving targets are located by using a projection inverse reconstruction method;

[0031] According to the one-dimensional projection curves T i (θ) of the scene image in the four directions of 0°, 45°, 90° and 135°, the binarization projection result B i (θ) is obtained:

[0032]

[0033] wherein ε is a threshold value;

[0034] The projection strips in the four directions are obtained by using the binarization projection result B i (θ), and the final combined projection result of the scene image of the moving target is obtained by merging:

[0035]

[0036] wherein is the projection strip of the j-th direction of the i-th scene image, j = 1, 2, 3, 4, corresponding to the four directions of 0°, 45°, 90° and 135° respectively;

[0037] The region where the projection strips in the four directions intersect is taken to obtain the region position A i (x,y) of the multiple moving targets in the i-th scene image of the moving target:

[0038]

[0039] According to the region position, an edge detection algorithm is used to obtain the number N of the moving targets and the boundary position in the i-th scene image of the moving target.

[0040] S22: region segmentation of the scene image according to the region position of the moving target;

[0041] By separating the moving targets in the horizontal and vertical directions, if there are N moving targets, the scene image is divided into N regions, so that there is only one moving target in each region.

[0042] Preferably, step S22: region segmentation of the scene image according to the region position of the moving target, specifically:

[0043] If the N moving targets can be completely separated in a horizontal or vertical single direction, the region segmentation is performed in the direction, and the position of the segmentation line is the midpoint line of the positions of the boundaries of two adjacent moving targets in the direction; if the N moving targets cannot be completely separated in a single direction, the region segmentation is performed in the direction in which the number of regions is the largest, and the position of the segmentation line is still the midpoint line of the positions of the boundaries of adjacent moving targets which can be separated, and then the region segmentation is performed in the other direction.

[0044] Preferably, the S3 modulates the scene image using the deformed geometric moment mode to obtain the position coordinates of the center of mass of each moving target, and specifically, the S3 comprises the following steps of:

[0045] S31: the method for generating the deformed geometric moment mode comprises the following steps of:

[0046] For a scene image with a size of MxN, the coordinate range of a moving target in the x direction after segmentation is (x1, x2), the coordinate range of the moving target in the y direction after segmentation is (y1, y2), the range length of the x direction segmentation region is L x =x2-x1+1, the range length of the y direction segmentation region is L y =y2-y1+1, and three gray deformed geometric moment modes S0, S1 and S2 are generated and expressed as:

[0047]

[0048]

[0049] S32: the scene image of the moving target is modulated using the deformed geometric moment mode to obtain a second modulation image;

[0050] S33: the position coordinates of the center of mass of the moving target are obtained;

[0051] The second light intensity value d0, the third light intensity value d1 and the fourth light intensity value d2 of the second modulation image are collected using a single-pixel detector:

[0052] The position coordinates (x', y') of the center of mass of the moving target relative to the segmentation region in which the moving target is located can be expressed as:

[0053]

[0054] For the entire scene image with a size of MxN and including the moving target, the position coordinates (x, y) of the center of mass of the moving target can be expressed as:

[0055]

[0056] According to S31, three gray scale deformation geometric moment patterns are generated for each moving target in the scene image, and then according to S32 and S33, the center coordinates (x, y) of each moving target in the scene image are obtained, which are the positions of the moving target.

[0057] Preferably, S4: obtaining the corresponding azimuth angle according to the center coordinates of each moving target, specifically:

[0058] S41: generating a second-order central moment pattern, the method is:

[0059] The center coordinates of a moving target Q in the scene image are (x Q ,y Q ), and the second-order central moment pattern S pq is described as:

[0060] S pq = (x-x Q ) p (y-y Q ) q ; (14)

[0061] wherein x Q and y Q are the center coordinates of the moving target Q, p+q=2, p∈{0,1,2}, q∈{0,1,2}; therefore the second-order central moment pattern S pq includes S 11 , S 20 and S 02 ;

[0062] The normalized difference second-order central moment pattern is obtained using the normalized difference method:

[0063]

[0064] wherein a and b are the minimum and maximum pixel gray scale values in S pq ;

[0065] S42: using the normalized difference second-order central moment pattern to modulate the scene image of the moving target to obtain a third modulation image;

[0066] S43: the azimuth angle of the moving target is obtained by:

[0067] Using a single-pixel detector to collect the fifth light intensity value and the sixth light intensity value

[0068] The second-order central moment of the moving target Q in the scene image is:

[0069]

[0070] Constructing the covariance matrix The two eigenvalues ​​λ1, λ2 of the covariance matrix T and the corresponding two eigenvectors v1, v2 are obtained; the azimuth angle θ of the moving target Q is:

[0071]

[0072] Among them, θ is the azimuth angle of the moving target Q, arctan() is the inverse tangent function, and v p (2) is the eigenvector v p The second component, v p (1) is the eigenvector v p The first component of

[0073] For each moving target in the scene image, S4 is used to obtain the azimuth angle of the corresponding moving target.

[0074] Preferably, in the step S41 of generating the second-order central moment pattern, the size of the second-order central moment pattern is 1.1 times the size of the moving target in the scene image.

[0075] Compared with the existing technology, the present invention has the following beneficial effects:

[0076] (1) The method for rapid detection of the position and azimuth of multiple moving targets based on single-pixel imaging of the present invention can simultaneously detect the position coordinates and azimuths of multiple moving targets, thereby achieving accurate and rapid detection.

[0077] (2) The position coordinates and azimuth angles of the moving target obtained by the present invention are accurate, and the number of modulations required for each frame is relatively small.

[0078] (3) By utilizing the high-speed modulation characteristics of DMD, the position and direction angle of the moving target can be detected quickly and accurately. BRIEF DESCRIPTION OF THE DRAWINGS

[0079] Figure 1 This is a flow chart of the method for rapidly detecting the positions and azimuths of multiple moving targets based on single-pixel imaging of the present invention;

[0080] Figure 2 This is a schematic diagram of one-dimensional projection results in four directions obtained for a frame of scene image according to the present invention;

[0081] Figure 3 Schematic diagram of the segmentation of multiple drone regions obtained from a certain frame of scene image according to the present invention;

[0082] Figure 4The schematic diagram of the deformed geometric moment mode modulated by the scene image of a frame according to the application;

[0083] Figure 5 The schematic diagram of the normalized differential second central moment mode modulated by the scene image of a frame according to the application;

[0084] Figure 6 The schematic diagram of the actual centroid trajectory and the centroid trajectory positioning result of three unmanned aerial vehicles according to the application;

[0085] Figure 7 The error result diagram of the centroid positioning of three unmanned aerial vehicles according to the application;

[0086] Figure 8 The comparison result diagram of the actual azimuth and the azimuth detection result of three unmanned aerial vehicles according to the application and the error result diagram of the azimuth detection of three unmanned aerial vehicles. DETAILED DESCRIPTION

[0087] According to the multi-motion target position and azimuth fast detection method based on single-pixel imaging, the motion target is set as an unmanned aerial vehicle, and the single-pixel system is used to measure the motion information of multiple unmanned aerial vehicles. The application will be further described in detail in combination with the drawings.

[0088] The multi-motion target position and azimuth fast detection method based on single-pixel imaging, as shown in Figure 1 the Fourier coefficients of each frame of scene image are obtained by using the single-pixel imaging system, and the one-dimensional projection curves of the scene in the directions of 0°, 45°, 90° and 135° are calculated; the region position of each unmanned aerial vehicle is preliminarily obtained by using the projection reverse reconstruction method, and the scene is segmented according to the region position of the unmanned aerial vehicle, that is, the scene image is divided into several regions if there are several unmanned aerial vehicles, so that there is only one unmanned aerial vehicle in each region; the scene image is modulated by using the deformed geometric moment mode, and the centroid position coordinates of each unmanned aerial vehicle can be obtained; the azimuth of each unmanned aerial vehicle is calculated by using the second central moment mode according to the centroid position coordinates of the unmanned aerial vehicle; specifically, the method comprises the following steps:

[0089] S1: for each frame of scene image including an unmanned aerial vehicle, the one-dimensional projection curve of the scene image is obtained.

[0090] For each frame of scene image including an unmanned aerial vehicle, the Fourier coefficients of the scene image are first obtained by using the single-pixel imaging system, and then the one-dimensional projection curves of the scene image in the directions of 0°, 45°, 90° and 135° are obtained; the specific steps are as follows:

[0091] S11: generate a series of different spatial frequencies and initial phases respectively, gray-scale Fourier pattern

[0092]

[0093] wherein a is the average light intensity value of the gray-scale Fourier pattern, b is the contrast of the gray-scale Fourier pattern, x and y are the spatial coordinates of the scene image respectively, u and v are the spatial frequencies of the discrete domain respectively; assuming that the size of the scene image of the UAV is M x N, the values of x are 1, 2, …, N respectively, the values of y are 1, 2, …, M respectively, the values of u are the values of v are is the initial phase, the values of are

[0094] S12: modulating the scene image of the UAV using the binary Fourier pattern.

[0095] The gray-scale Fourier pattern is converted into a binary Fourier pattern using a Floyd-Steinberg error dithering algorithm and the binary Fourier pattern is loaded by a DMD (Digitial Micromirror Devices) to modulate the i-th scene image of the UAV to obtain a first modulated image of the i-th scene image.

[0096] S13: collecting the total light intensity value of the first modulated image using a single-pixel detector.

[0097] For the gray-scale Fourier pattern with the initial phase with spatial frequencies u and v, the total first light intensity value of the first modulated image is collected using a single-pixel detector :

[0098]

[0099] wherein e is the background noise, I i (x, y) is the i-th scene image; is the binary Fourier pattern, is the first light intensity value.

[0100] S14: calculating the Fourier coefficient F i (u, v) at the frequency (u, v) using a three-step phase shift method formula

[0101]

[0102] According to S11 to S14, for the i-th frame of scene image of the UAV, first, modulation is performed using S12 according to a series of gray Fourier modes in S11, then the initial phase is used to obtain the i-th frame of scene image according to formula (2) The first light intensity value after modulation of the gray Fourier mode with spatial frequency (u, v), and finally, formula (3) is used to obtain the Fourier coefficient F i (u, v) of the i-th frame of scene image of the UAV at frequency (u, v).

[0103] S15: Obtain the one-dimensional projection curve of the scene image in the four directions of 0°, 45°, 90° and 135°.

[0104] Based on the Fourier slice theorem: the Fourier transform result of the one-dimensional projection curve of a two-dimensional image in a certain direction is equivalent to the slice result of the two-dimensional Fourier transform of the image with the origin in the same direction; therefore, the one-dimensional projections of the scene in the four directions of 0°, 45°, 90° and 135° are equivalent to the one-dimensional inverse Fourier transform results of the slices in the corresponding 0°, 45°, 90° and 135° directions of the two-dimensional Fourier transform of the image with the origin:

[0105]

[0106] where θ is the included angle between the projection direction and the x-axis direction, T i (θ) is the projection curve of the i-th frame of scene image in the θ direction, F i -1 is the one-dimensional inverse Fourier transform of the Fourier coefficient F i (u, v).

[0107] S2: Divide the scene image into N regions using one-dimensional projection.

[0108] The projection reverse reconstruction method is used to preliminarily obtain the regions where the multiple UAVs are located, and the scene image is regionally segmented according to the region positions of the UAVs; if there are several UAVs, the scene image is divided into several regions, so that only one UAV exists in each region.

[0109] S21: The projection reverse reconstruction method is used to preliminarily obtain the regions where the multiple UAVs are located, and the specific method is as follows:

[0110] Since there is a certain contrast between the UAV and the background, the one-dimensional projection value of the region where the UAV is located is relatively high, and the one-dimensional projection value of the region where the UAV does not exist is relatively low; a threshold ε is set to process the projection curve T i (θ) to obtain the binary projection result B i (θ):

[0111]

[0112] According to step S1 and expression (4), the one-dimensional projection curve of the scene in four directions of 0°, 45°, 90° and 135° is used in this embodiment, and thus the projection strips of the i-th frame of the scene image are obtained by using the inverse reconstruction of projection according to expression (5) corresponding to the four directions of 0°, 45°, 90° and 135° in expression (4). The projection strips in the four directions are combined to obtain the final combined projection result of the scene image of the UAV

[0113]

[0114] The region position A of the multiple UAVs in the i-th frame of the scene image of the UAV is obtained by taking the intersection of the projection strips in the four directions i (x, y):

[0115]

[0116] According to the region position, the number N of the UAVs and the boundary position in the i-th frame of the scene image are obtained by using an edge detection algorithm.

[0117] S22: The specific method of region segmentation of the scene image according to the region position of the UAV is as follows:

[0118] If the N UAVs can be completely separated in a single horizontal or vertical direction, the region segmentation is preferentially performed in the direction, and the position of the segmentation line is the midpoint line of the boundary positions of the two adjacent UAVs in the direction; if the N UAVs cannot be completely separated in a single direction, the region segmentation is preferentially performed in the direction that can divide the most regions, and the position of the segmentation line is still the midpoint line of the boundary positions of the adjacent UAVs that can be divided, and then the region segmentation is performed in another direction, and the scene image is divided into N regions so that only one UAV exists in each region.

[0119] S3: The centroid position coordinates of each UAV are obtained by modulating the scene image of the UAV using the deformed geometric moment mode.

[0120] S31: The method of generating the deformed geometric moment mode is as follows:

[0121] For a scene image with a size of MxN, assuming that the coordinate range of the x-direction segmentation region of a certain UAV after segmentation is (x1, x2), the coordinate range of the y-direction segmentation region is (y1, y2), the length of the x-direction segmentation region range is L x =x2-x1+1, and the length of the y-direction segmentation region range is L y= y2 - y1 + 1, three gray scale morphing geometric moment patterns S0, S1, S2 are generated and expressed as:

[0122]

[0123] wherein AND represents the AND relationship, taking expression (8) as an example, when i belongs to the coordinate range (x1, x2) of the x-direction partitioned region and j belongs to the coordinate range (y1, y2) of the y-direction partitioned region, the value of element a ji in the gray scale morphing geometric moment pattern S0 is 1; when i does not belong to the coordinate range (x1, x2) of the x-direction partitioned region or j does not belong to the coordinate range (y1, y2) of the y-direction partitioned region, the value of element a ji in the gray scale morphing geometric moment pattern S0 is 0.

[0124] S32: Modulating the scene image of the UAV using the morphing geometric moment pattern.

[0125] The three gray scale morphing geometric moment patterns are respectively converted into three binary morphing geometric moment patterns S'0, S'1, S'2 by using the Floyd-Steinberg error dithering algorithm, and the second modulated image is obtained by modulating the i-th frame of scene image of the UAV using the generated binary morphing geometric moment pattern by using the DMD.

[0126] For the traditional geometric moment method, a gray scale geometric moment pattern is used, wherein the gray scale level of the first moment pattern is related to the image size, for example, if the size of the scene image is 256x256, the gray scale level of the traditional first moment pattern is 256. Therefore, the error of the position of the barycenter obtained after the existing gray scale geometric moment pattern is dithered and binarized is large; here, the gray scale morphing geometric moment pattern is used to reduce the gray scale level of the first moment pattern: the gray scale level of the first moment pattern in the x-direction is L x , and the gray scale level of the first moment pattern in the y-direction is L y After the gray scale morphing geometric moment pattern is dithered and binarized, the error of the position of the barycenter is reduced compared with the traditional method.

[0127] S33: The method for obtaining the position coordinates of the barycenter of each UAV is:

[0128] The total light intensity value of the second modulated image is collected by using a single-pixel detector, and the corresponding total light intensity values after the i-th frame of scene image is respectively modulated by using the binary morphing geometric moment patterns S'0, S'1, S'2 are respectively the second light intensity value d0, the third light intensity value d1 and the fourth light intensity value d2:

[0129]

[0130] wherein I i(x, y) is the i-th frame of the scene image.

[0131] The coordinates of the center of mass (x', y') of the UAV relative to the partition region in which it is located can be expressed as:

[0132]

[0133] The coordinates of the center of mass (x, y) of the UAV in the entire MxN size scene image including the UAV are expressed as:

[0134]

[0135] For each UAV in the scene image, three gray-scale deformation geometric moment patterns are generated according to S31, and then the coordinates of the center of mass (x, y) of each UAV in the scene image are obtained according to S32 and S33. The coordinates of the center of mass (x, y) of the UAV in the scene image are the position of the UAV.

[0136] S4: Obtain the azimuth angle according to the coordinates of the center of mass of the UAV.

[0137] According to the coordinates of the center of mass of the UAV in S3, the second-order central moment pattern is modulated for each UAV, and the azimuth angle of each UAV is calculated, which is specifically:

[0138] S41: Generate the second-order central moment pattern, the method is:

[0139] The coordinates of the center of mass of each UAV are obtained by step S3, and the second-order central moment pattern is modulated for each UAV; assuming that the coordinates of the center of mass of a certain UAV Q are (x Q ,y Q ), the second-order central moment pattern S pq can be described as:

[0140] S pq =(x-x Q ) p (y-y Q ) q ; (14)

[0141] Wherein, x Q and y Q are the coordinates of the center of mass of the UAV Q, which are obtained according to step S3, p+q=2, p∈{0,1,2}, q∈{0,1,2}; considering the permutation and combination of p and q, the second-order central moment pattern S pq includes S 11 , S 20 and S 02 .

[0142] To reduce the binarization error, the size of the second-order central moment pattern can be considered to be reduced, so as to reduce the gray scale; knowing the size of the UAV in the scene image, the size of the second-order central moment pattern can be selected to be slightly larger than the size of the UAV in the scene image, for example, the size of the second-order central moment pattern is selected to be 1.1 times the size of the UAV, and the size of the UAV in the scene image is 100pixelx100pixel, so the size of the second-order central moment pattern is 110pixelx110pixel in a range centered on x Q and y Q . In addition, in order to reduce the noise caused by the background light, the normalized difference method is used to obtain the normalized difference second-order central moment pattern:

[0143]

[0144] wherein a and b are the minimum and maximum pixel gray values in S pq , S pq+ and S pq- satisfy:

[0145]

[0146] S42: using the normalized difference second-order central moment pattern to modulate the scene image of the UAV.

[0147] The Floyd-Steinberg error dithering algorithm is used to convert the normalized difference second-order central moment pattern into a binary normalized difference second-order central moment pattern S′ pq+ and S′ pq- , and the DMD is used to load the binary normalized difference second-order central moment pattern S′ pq+ and S′ pq- to modulate the scene image of the UAV. Because each UAV in the scene image corresponds to three second-order central moment patterns S pq , S 11 , S 20 and S 02 , and each of them uses the normalized difference method to obtain a normalized difference second-order central moment pattern, therefore, for each UAV in the i-th scene image, six central moment patterns are used for modulation to obtain a third modulation image, and if there are N UAVs in the scene image, 6N times of modulation are required.

[0148] S43: the azimuth angle calculation method of the UAV is:

[0149] After modulation by the binary normalized difference second-order central moment pattern S′ pq+ and S′ pq- , the total light intensity value of the third modulation image collected by the single-pixel detector includes the fifth light intensity value and the sixth light intensity value for:

[0150]

[0151] Where, e is the background noise, I i (x,y) is the i-th frame scene image.

[0152] Similarly, taking the drone Q in the scene image as an example, its second-order central moment is:

[0153]

[0154] According to the relationship between the second-order central moment mode and the second-order central moment mode of the normalized difference, as well as the relationship between the fifth light intensity value, the sixth light intensity value and the second-order central moment mode of the normalized difference, it can be seen that the second-order central moment can be obtained from the fifth light intensity value of the third modulated image. and the sixth light intensity value Get as follows:

[0155]

[0156] Constructing the covariance matrix Assume that the two eigenvalues ​​of the covariance matrix T are λ1 and λ2, and the corresponding two eigenvectors are v1 and v2; the larger eigenvalue of λ1 and λ2 is λ p , the corresponding eigenvector is v p ;Right now Through principal component analysis theory, the azimuth angle θ of the UAV can be obtained from the eigenvector v of the covariance matrix T p Calculation yields:

[0157]

[0158] Where θ is the azimuth angle of the UAV Q, v p (2) is the eigenvector v p The second component, v p (1) is the eigenvector v p The first component of .

[0159] For each UAV in the scene image, S4 is used to obtain the azimuth angle of the corresponding UAV.

[0160] The embodiment of the present application specifically provides a multi-unmanned aerial vehicle position and azimuth detection method based on single-pixel imaging, which utilizes a single-pixel imaging system to obtain motion information of multiple moving unmanned aerial vehicles. A scene image is 512pixel*512pixel, wherein the scene image comprises three unmanned aerial vehicles, the unmanned aerial vehicles are all gray scale images, and the gray scale values are different. The sizes of the three unmanned aerial vehicles are 60pixel*60pixel, 70pixel*70pixel and 80pixel*80pixel respectively, one side of the unmanned aerial vehicle performs a translation motion, the motion trajectories are infinity, a circle and an ellipse respectively, one side of the unmanned aerial vehicle performs a random rotation motion, and the scene image has a total of 360 frames. The present application simultaneously measures the position coordinates and the azimuth of the three unmanned aerial vehicles for each frame of the scene image.

[0161] S1: for each frame of the scene image, a one-dimensional projection curve of the scene image is obtained.

[0162] For each frame of the scene image, first, the Fourier coefficients of the scene are obtained by using the single-pixel imaging system, and the one-dimensional projection curves of the scene image in the directions of 0°, 45°, 90° and 135° are calculated; the one-dimensional projection curves of the scene image in the directions of 0°, 45°, 90° and 135° are equal to the one-dimensional inverse Fourier transform results of the corresponding 0°, 45°, 90° and 135° direction slices of the two-dimensional Fourier transform of the scene image through the origin:

[0163]

[0164] wherein θ is the included angle between the projection direction and the x-axis direction, T i (θ) is the projection curve of the i-th frame of the scene image in the direction of θ, F i -1 is the one-dimensional inverse Fourier transform of the Fourier coefficient F i (u, v) of the scene image.

[0165] Therefore, here, only the Fourier coefficients at the frequencies (u, v) on the straight lines passing through the origin and having the included angles of 0°, 45°, 90° and 135° with the x-axis need to be sampled.

[0166] The present application does not need to be fully sampled. Assuming that the sampling rate is η, when η is 0.05, for a frame of the scene image of 512pixel*512pixel, 512*0.05≈26 frequency points (including the common origin) are sampled in each direction, the common origin is removed, that is, 25 Fourier coefficients are sampled in each direction, according to the three-step phase shift method, each Fourier coefficient needs to project 3 patterns, and there are 4 directions, that is, 25*3*4 patterns, since the Fourier coefficient at the origin is a real value, it only needs to be sampled twice, therefore, each frame of the scene image needs to project P1=25*3*4+2=302 patterns in step S1. For example, if the scene image is 512pixel*512pixel, the sampling rate is 0.05, and the scene image has 360 frames, the total number of patterns needed to be projected is P2=360*302=108,720.Figure 2 It is a schematic diagram showing the one-dimensional projection results in four directions obtained by executing step S1 on a certain frame of scene image according to the present invention. Figure 2 It only schematically shows three projections of three initial phases of a spatial frequency, which does not mean that there are only three projections in one direction.

[0167] S2: Divide the scene image into N regions using one-dimensional projection.

[0168] The projection inverse reconstruction method is used to preliminarily obtain the areas where multiple drones are located, and the scene image is segmented according to the location of the drone area. The scene image is divided into several areas according to the number of drones, so that there is only one drone in each area; Figure 3 As shown in FIG, a schematic diagram of multiple drone region segments obtained by executing step S2 on a certain frame of scene image in the present invention.

[0169] S3: The scene image of the UAV is modulated using the deformed geometric moment pattern to obtain the center of mass position coordinates of each UAV.

[0170] like Figure 4 The figure shows a schematic diagram of the deformation geometric moment pattern modulated by the present invention for a frame of scene image. Each drone needs to project three deformation geometric moment patterns. For the case of three moving drones, each frame of scene image needs to project P2 = 9 deformation geometric moment patterns in step S3.

[0171] Find the center of mass position of each UAV. Figure 6 As shown in FIG, the center-of-mass motion trajectory positioning results and actual center-of-mass trajectory diagram of the three UAVs of the present invention are shown in FIG. Figure 7 Figure 2 shows the error results of the center of mass positioning of three drones using the present invention. The results show that the center of mass error for most frames of drones 2 and 3 in both the x and y directions is between -2 and 2 pixels. Occasionally, the absolute value of the angular error is slightly greater than 2 pixels, but never exceeds 3 pixels. The results for drone 1 in the x direction are similar to those of drones 1 and 2. In the y direction, the center of mass error for most frames is between -3 and 3 pixels, with a few frames showing an absolute value slightly greater than 3 pixels, but never exceeding 5 pixels. This demonstrates that the center of mass positioning results are relatively accurate.

[0172] The root mean square error (RMSE) is used to evaluate the positioning accuracy. Assume that the actual center of mass coordinates of a certain drone in the i-th frame is The detected center of mass coordinates are There are K frames of scene images in total, so the root mean square error of the drone positioning result is:

[0173]

[0174] Through calculation, the root mean square error results of the centroid positioning of the unmanned aerial vehicles 1-3 are 1.418, 0.996 and 0.995 respectively, and it can be seen that the centroid positioning result of the present application is relatively accurate.

[0175] S4: according to the centroid coordinates of the unmanned aerial vehicle in S3, each unmanned aerial vehicle is modulated by using a second-order central moment mode, and the azimuth angle of each unmanned aerial vehicle is calculated.

[0176] As shown in Figure 5 , it is a schematic diagram of the normalized difference second-order central moment mode for modulating a certain frame of scene image according to the present application. Each unmanned aerial vehicle needs to project 6 difference second-order central moment modes, and for the case of 3 moving unmanned aerial vehicles, each frame of scene step S4 needs to project P3=18 modes. Therefore, a total of P=P1+P2+P3=329 modes need to be projected for each frame of scene.

[0177] As shown in Figure 8 , it is a comparison result diagram of the azimuth angle detection result of the three unmanned aerial vehicles according to the present application and the actual azimuth angle, and an error result diagram of the azimuth angle detection of the three unmanned aerial vehicles. As can be seen from Figure 8 , the azimuth angle error of the unmanned aerial vehicle 3 is between-5° and 5°; the azimuth angle error of the unmanned aerial vehicle 2 is mostly between-5° and 10°, and the angle error of two frames exceeds 10°; the azimuth angle error of the unmanned aerial vehicle 1 is mostly between-5° and 15°, and the angle error of one frame slightly exceeds 15°. Similarly, the root mean square error of the angle measurement is used to evaluate the accuracy of the measurement: assuming that the actual azimuth angle of the i-th frame of a certain unmanned aerial vehicle is θ i , the detected azimuth angle is β i , and the scene image has K frames, then the root mean square error RMSE θ of the azimuth angle detection result of the unmanned aerial vehicle is defined as:

[0178]

[0179] Through calculation, the root mean square error results of the azimuth angles of the unmanned aerial vehicles 1-3 are 6.29, 3.46 and 1.74 respectively, and it can be seen that the azimuth angle result of the present application is relatively accurate.

[0180] The present application has the following beneficial effects: the present application can realize the rapid detection of the position coordinates and azimuth angles of multiple moving unmanned aerial vehicles; the position coordinates and azimuth angle results of the moving unmanned aerial vehicles calculated by the present application are relatively accurate, and the number of detection speckles (i.e. modes) required by each frame is relatively small; the high-speed modulation characteristics of the DMD can be used to realize the rapid and high-precision detection of the position and direction angles of the unmanned aerial vehicles, which is convenient for further controlling the flight direction and path of the unmanned aerial vehicles.

[0181] The above-described embodiments are merely intended to describe the preferred embodiments of the present application, and are not intended to limit the scope of the present application. Various changes and modifications made by those skilled in the art to the technical solutions of the present application without departing from the design spirit of the present application shall fall within the scope of protection of the present application.

Claims

1. A method for fast detection of position and azimuth angle of multiple moving targets based on single-pixel imaging, characterized in that, It comprises the following steps: S1: for each frame of scene image comprising a plurality of moving targets, obtaining a one-dimensional projection curve of the scene image; For each frame of the scene image, the first light intensity value is obtained by modulating the scene image with Fourier mode, and the Fourier coefficient F i (u,v) is obtained according to the first light intensity value by using the three-step phase shift formula i (u,v) is the Fourier coefficient of the i-th frame of the scene image at the frequency (u,v); and the one-dimensional projection curve of the scene image in the four directions of 0°, 45°, 90° and 135° is obtained according to the Fourier coefficient F i (u,v). S2: dividing the scene image into N regions using one-dimensional projection; Using the projection inverse reconstruction method to obtain the region where the N moving targets are located, and according to the region position of the moving target, the scene image is regionally segmented, and the scene image is divided into N regions when there are N moving targets, so that there is only one moving target in each region; wherein N is a positive integer; S3: using the deformed geometric moment mode to modulate the scene image to obtain the centroid coordinate of each moving target; According to the coordinate range of the region where each moving target is located, the corresponding deformed geometric moment mode is generated, and the second, third and fourth light intensity values are obtained after the scene image is modulated using the deformed geometric moment mode. According to the second, third and fourth light intensity values, the centroid coordinates of the moving target relative to the region where it is located are obtained, and then converted into the centroid coordinates of the moving target in the scene image; S4: obtaining the corresponding azimuth angle according to the centroid coordinates of each moving target; According to the centroid coordinates of the moving target in S3, a second-order central moment mode is obtained, a normalized difference second-order central moment mode is obtained by using a normalized difference method on the second-order central moment mode, and fifth and sixth light intensity values are obtained by modulating the scene image using the normalized difference second-order central moment mode. According to the fifth and sixth light intensity values, the second-order central moment is obtained, a covariance matrix is constructed according to the second-order central moment, two eigenvalues and corresponding two eigenvectors of the covariance matrix are obtained, and the azimuth angle of the moving target is obtained according to the eigenvector.

2. The method of claim 1, wherein, In step S1, for each frame of scene image comprising a moving target, the one-dimensional projection curve of the scene image is obtained as follows: S11: generating a series of different spatial frequencies and initial phases respectively gray-scale Fourier patterns Wherein, a is the average light intensity value of the gray Fourier mode, b is the contrast of the gray Fourier mode, x, y are the spatial coordinates of the scene image respectively, u, v are the spatial frequency of the discrete domain respectively; when the size of the scene image of the moving target is MxN, the value of x is 1, 2, …, N respectively, the value of y is 1, 2, …, M respectively, the value of u is The value of v is The value of v is S12: using a binary Fourier mode to modulate the scene image of the moving target to obtain a first modulated image; S13: using a single-pixel detector to collect the first light intensity value of the first modulated image; wherein e is background noise, I i (x,y) is the i-th frame of the scene image; is a binary Fourier mode, is a first light intensity value; S14: Calculate the Fourier coefficients F at frequencies (u, v) using the three-step phase shift method formula i (u,v): S15: obtaining the one-dimensional projection curve of the scene image in the directions of 0°, 45°, 90° and 135°; The one-dimensional projection of the scene in the directions of 0°, 45°, 90° and 135° is equal to the one-dimensional inverse Fourier transform result of the corresponding 0°, 45°, 90° and 135° direction slice of the two-dimensional Fourier transform over the origin of the image: where θ is the angle between the projection direction and the x-axis direction, T i (θ) is the projection curve of the i-th frame of scene image in the θ direction, F i -1 is the Fourier coefficient F i is the one-dimensional inverse Fourier transform of F 3. The method of claim 1, wherein, In step S2, the scene image is divided into N regions using one-dimensional projection, specifically as follows: S21: preliminarily obtaining the regions where the plurality of moving targets are located using the projection inverse reconstruction method; According to the one-dimensional projection curve T of the scene image in the four directions of 0°, 45°, 90° and 135° i (θ) is processed to obtain a binary projection result B i (θ): Wherein, ε is a threshold value; Using the binarized projection result B i (θ) to obtain four projection strips in different directions, and to obtain the final combined projection result of the scene image of the moving target by merging wherein, is the projection strip of the i-th frame scene image in the j-th direction, j = 1, 2, 3, 4, corresponding to the four directions of 0°, 45°, 90°, 135°, respectively; Taking the region where the projection strips in four directions intersect, the region position A of the multiple moving targets in the i-th frame scene image of the moving target is obtained i (x, y): According to the region position, an edge detection algorithm is used to obtain the number N of moving targets and the boundary position in the i-th frame of scene image; S22: regionally segmenting the scene image according to the region position of the moving target; By separating the moving targets in the horizontal and vertical directions, the scene image is divided into N regions when there are N moving targets, so that there is only one moving target in each region.

4. The method of claim 3, wherein, In step S22, the scene image is regionally segmented according to the region position of the moving target, specifically as follows: If the N moving targets can be completely separated in a horizontal or vertical single direction, the region segmentation is performed in the direction, and the position of the segmentation line is the midpoint line of the positions of the boundaries of two adjacent moving targets in the direction; if the N moving targets cannot be completely separated in a single direction, the region segmentation is performed in the direction in which the number of regions is the largest, and the position of the segmentation line is still the midpoint line of the positions of the boundaries of adjacent moving targets that can be separated, and then the region segmentation is performed in another direction.

5. The method of claim 1, wherein, The S3 modulates the scene image using the deformed geometric moment mode to obtain the position coordinates of the center of mass of each moving target, and specifically comprises the following steps: S31: The method for generating the deformed geometric moment mode comprises the following steps: For a scene image of size M x N, the coordinate range of the x direction segmentation region of a certain moving target after segmentation is (x1, x2), the coordinate range of the y direction segmentation region is (y1, y2), the length of the x direction segmentation region range is L x =x2-x1+1, the length of the y direction segmentation region range is L y =y2-y1+1, then three gray level deformation geometric moment patterns S0, S1 and S2 are generated and expressed as: S32: The scene image of the moving target is modulated using the deformed geometric moment mode to obtain a second modulation image; S33: The position coordinates of the center of mass of the moving target are obtained; The second light intensity value d0, the third light intensity value d1 and the fourth light intensity value d2 of the second modulation image are collected using a single-pixel detector: The position coordinates (x', y') of the center of mass of the moving target relative to the segmented region in which the moving target is located can be represented as: For the entire scene image of the size of MxN including the moving target, the position coordinates (x, y) of the center of mass of the moving target are represented as: For each moving target in the scene image, three gray deformed geometric moment modes are generated according to S31, and then the position coordinates (x, y) of the center of mass of each moving target in the scene image are obtained according to S32 and S33. The position coordinates (x, y) of the center of mass of the moving target in the scene image are the position of the moving target.

6. The method of claim 1, wherein, The S4: The corresponding azimuth angle of each moving target is obtained according to the position coordinates of the center of mass, and specifically comprises the following steps: S41: A second-order central moment mode is generated, and the method comprises the following steps: The centroid coordinates of a moving target Q in a scene image are (x Q ,y Q ), and the second-order central moment mode S pq is described as: S pq = (x - x Q ) p (y - y Q ) q ; (14) where x Q and y Q are the center of mass coordinates of the moving target Q, p+q=2, p∈{0,1,2}, q∈{0,1,2}; thus the second order central moment pattern S pq includes S 11 , S 20 and S 02 three; A normalized differential second-order central moment mode is obtained using a normalized differential method: where a and b are the minimum and maximum pixel gray values in the image, respectively pq where a and b are the minimum and maximum pixel gray values in the image, respectively S42: The scene image of the moving target is modulated using the normalized differential second-order central moment mode to obtain a third modulation image; S43: The method for obtaining the azimuth angle of the moving target comprises the following steps: acquiring a fifth light intensity value of the third modulated image using the single-pixel detector and a sixth light intensity value The second-order central moment of the moving target Q in the scene image is: Constructing a Covariance Matrix Two eigenvalues λ1, λ2 and two corresponding eigenvectors v1, v2 of the covariance matrix T are obtained; the azimuth angle θ of the moving target Q is: Among them, θ is the azimuth angle of the moving target Q, arctan() is the inverse tangent function, and v p (2) is the eigenvector v p The second component, v p (1) is the eigenvector v p The first component of The azimuth angle of each moving target in the scene image is obtained using S4.

7. The method of claim 6, wherein the method is a single-pixel imaging based multi-moving target position and azimuth angle fast detection method. In the S41, the size of the second-order central moment mode is 1.1 times the size of the moving target in the scene image.

8. The method of claim 1-7, wherein, The moving target is a UAV.

Images