Target 3D reconstruction method based on multi-view radar images

Abstract

A method for three-dimensional reconstruction of a target based on multi-view radar images includes: step 1, receiving echo signals within a set time period by an ISAR radar system; performing range-Doppler processing on the echo signals within the set time period to obtain a radar image sequence of the target area, the sequence including P radar images of the target area; step 2, using a dual-parameter constant false alarm algorithm to distinguish target pixels from background pixels to obtain a binary segmented image; using methods such as morphological operators to extract the target contour; step 3, performing centroid matching on the ISAR image using a parameter search method; and step 4, constructing the target voxels using a spatial projection method. The present invention utilizes processing techniques such as constant false alarm segmentation, centroid matching, and visual surface visualization to achieve three-dimensional visualization reconstruction of the target and extract target posture information. Contour feature extraction is used to extract and correlate the multi-view target image sequence, enhancing the robustness of feature extraction.

Description

Technical Field

[0001] The present invention belongs to the field of radar technology, and in particular relates to a target three-dimensional reconstruction method based on multi-view radar images, which can be used to reconstruct space targets in three dimensions using a single-station multi-view ISAR image sequence. Background Art

[0002] Space target attitude estimation is crucial for understanding a space target's motion intent and determining its state. Using sequential images obtained from continuous measurements of space targets using Inverse Synthetic Aperture Radar (ISAR) to accurately determine the absolute attitude of key payload components, such as solar panels and flat-panel antennas, is a practical technology currently available for space target attitude estimation, with applications in civil and military fields, including space target fault analysis and threat assessment.

[0003] Currently, two common methods for measuring the attitude of space targets are active and passive. Active measurement involves installing a GPS receiver and inertial measurement unit (IMU) on an aircraft to measure its position and attitude. This method, known as moving target internal measurement, requires the aircraft to be equipped with a specialized measurement system, resulting in high system complexity. Passive measurement utilizes optoelectronic or radar sensors to continuously perform high-resolution imaging of space targets and determine their attitude based on the imaging sequence information. This passive measurement method has the advantage of being independent of the state of the space target and can also be directly applied to attitude measurement of unknown or out-of-control targets. However, using optoelectronic sensors to measure the attitude of space targets is susceptible to factors such as weather and time, and thus has limitations in actual space target attitude measurement. Summary of the Invention

[0004] In view of the above-mentioned deficiencies in the prior art, the present invention proposes a method for three-dimensional reconstruction of a target from multi-view radar images.

[0005] In order to achieve the above technical objectives, the present invention adopts the following technical solutions.

[0006] A method for three-dimensional reconstruction of a target based on multi-view radar images, characterized by comprising the following steps:

[0007] Step 1: The ISAR radar system receives echo signals within a set time period; performs range-Doppler processing on the echo signals within the set time period to obtain a radar image sequence of the target area, the sequence including P radar images of the target area;

[0008] Step 2: Use a dual-parameter constant false alarm algorithm to distinguish target pixels from background pixels to obtain a binary segmentation image; use morphological operators and other methods to extract the target contour;

[0009] Step 3, use parameter search method to perform centroid matching on ISAR images;

[0010] Step 4: Use the spatial projection method to construct the target voxel.

[0011] The present invention calculates the radar imaging projection matrix corresponding to the LOS sequence of the obtained multi-view ISAR radar image, then uses digital image processing technology to extract the target contour information of the image sequence, and then constructs an initial solid composed of several voxels in space. Finally, a projection method is used to determine whether each voxel falls into the target contour area after being projected onto the projection plane of each frame imaging, so as to realize the target three-dimensional surface reconstruction.

[0012] The advantages of the present invention compared with the prior art are:

[0013] 1) The present invention utilizes constant false alarm segmentation, centroid matching, and surface visualization processing technologies to achieve three-dimensional visualization reconstruction of the target and extraction of target posture information;

[0014] 2) The present invention uses contour feature extraction for the first time to extract and associate multi-view target image sequences, thereby enhancing the robustness of feature extraction. BRIEF DESCRIPTION OF THE DRAWINGS

[0015] Figure 1 is a flow chart of the present invention;

[0016] Figure 2 This is a 3D model diagram of the Tiangong-1 target used in the simulation of the present invention;

[0017] Figure 3a 、 Figure 3b 、 Figure 3c and Figure 3d These are the Tiangong-1 contour extraction results for the 1st to 4th frames of the ISAR image sequence; the horizontal axis is the azimuth distance (unit: m), and the vertical axis is the vertical distance (unit: m); Figure 3e This is the 3D reconstruction result of Tiangong-1;

[0018] Figure 4 It is a schematic diagram of the ISAR observation geometry of space targets;

[0019] Figure 5a 、 Figure 5b 、 Figure 5c 、 Figure 5d They are simulated Tiangong-1 ISAR images from different viewing angles;

[0020] Figure 6a is the three-dimensional visual shell reconstructed by the present invention;

[0021] Figure 6bThis is a comparison chart between the reconstruction result and the real 3D visual. DETAILED DESCRIPTION

[0022] Reference Figure 1 The method for 3D target reconstruction based on multi-view radar images of the present invention is specifically implemented in the following steps:

[0023] Step 1: The ISAR radar system receives echo signals within a set time period; the echo signals within the set time period are processed using a range-Doppler algorithm to obtain a radar image sequence of the target area, the sequence including P radar images of the target area;

[0024] Step 2: Use a dual-parameter constant false alarm method to perform background segmentation and contour extraction on the radar image;

[0025] The specific sub-steps of step 2 are:

[0026] 2.1 Calculate the mean and variance of all pixels in the image;

[0027] 2.2 Based on the calculated mean and variance, each pixel is divided into two categories: target and background according to the following formula;

[0028] Threshold=μ+a1σ (1)

[0029] Where Threshold is the threshold, μ is the mean of pixels in the image, σ is the variance of pixels in the image, and a1 is the coefficient.

[0030] 2.3 Based on the classification results, calculate the mean and variance of background pixels;

[0031] 2.4 Based on the mean and variance in step 2.3, classify the background pixels again according to the following formula:

[0032] Threshold'=μ'+a2σ′ (2)

[0033] Where Threshold′ is the threshold, μ′ is the mean of the pixels in the image in step 2.3, σ′ is the variance of the pixels in the image in step 2.3, and a2 is the coefficient.

[0034] 2.5 Calculate the mean and variance of the background pixels and compare them with the last classification threshold. If the difference between the two is less than the given threshold, stop the iteration, otherwise return to step 2.3.

[0035] Step 3: Match the target centroid and determine the voxel position;

[0036] Step 4, construct the target voxel using the spatial projection method;

[0037] The specific sub-steps of step 4 are:

[0038] 4.1 Calculate the two-dimensional contour set G i The distance between them and the point Is it beyond this distance? If it is, the point is not considered to fall within the target area and output 0, otherwise go to step 4.2;

[0039] Among them G i A set of two-dimensional contours consisting of edge points constitute, is a point in the radar image;

[0040] 4.2 Calculate the vector of each point and

[0041] 4.3 Determine the quadrant number Qu(j) of each edge point based on the point vector, which is 0 to 3;

[0042] 4.4 Determine the vectors of two points and The symbol for the cross product is:

[0043]

[0044] The factor ε i =max{|r1|…|r j ||d1|…|d1|}×6.6613×10 -6 ;

[0045] 4.5 Calculate the quadrant difference DQu(j) = Qu(j) - Qu(j+1) of each edge point and update it according to the following criteria:

[0046]

[0047] 4.6 Calculate the sum of the quadrant difference sequence. If the sum is not 0, the point is determined to be a point within the target and the output is 1. Otherwise, it is not a point within the target and execute step 4.7.

[0048] 4.7 Determining Point Vectors and The symbol of the inner product between

[0049]

[0050] If there exists j such that Sc(j)=0 and at the same time Inn(j)<0, the point is considered to be on the target contour and also considered to be a point inside the target, and the output is 1; otherwise the output is 0.

[0051] The effects of the present invention can be further illustrated by the following simulation experiments:

[0052] 1) Simulation conditions:

[0053] The 3D model of the typical space target Tiangong-1 (TG-I) used in the simulation of the present invention is as follows: Figure 2 The main parameters of the simulation of the present invention are shown in Table 1:

[0054] Table 1 Main parameters of ISAR radar system

[0055] ISAR image size 512×512 Wavelength λ 0.016m Bandwidth of the transmitted signal B 1GHz <![CDATA[Center frequency f of the transmitted signal c > 18.7GHz Pulse repetition frequency prf 100Hz

[0056] 2. Simulation content and result analysis:

[0057] Simulation 1: Using the method of the present invention Figure 2 The contour of the Tiangong-1 target is extracted, and the continuous target contours are as follows Figure 3(a) 、 3(b) As shown in Figures 3(c) and 3(d), based on the target's contour information and LOS parameter sequence, the method of the present invention is applied to perform three-dimensional reconstruction of the target. Its visual surface is shown in Figure 3(e). Compared with the Tiangong-1 model, the reconstruction result reflects most of the target's structural features, proving the effectiveness of the invention.

[0058] Simulation 2: The target reconstruction results of a single-station ground-based radar space target with limited observation angle change are simulated using the method of the present invention. Figure 4 As shown, the target contour is segmented using the method of the present invention. Figure 5(a) 、 5(b) , 5(c), 5(d), the 3D reconstructed shell and the comparison results with the real model are shown in Figure 6(a) and 6(b) As shown;

[0059] 3. Analysis of simulation results:

[0060] It can be seen from Figure 3(a), Figure 3(b) and Figure 3(c) that the extraction of the linear structure of the solar wing boundary of the space target is basically correct, but the non-solar wing part of the space target affects the accuracy of the linear structure extraction of the solar wing boundary to a certain extent.

[0061] As can be seen from FIG5(a), FIG5(b) and FIG5(c), the use of the change information of the three linear structures can well complete the association of the linear structures between the images after the boundaries are extracted.

[0062] As can be seen from Table 2, the orientation of the estimated linear structure is basically consistent with the orientation of the true linear structure, with an average error within 1 degree (0.63 degrees). In addition, it can be found that linear structure 1 is basically parallel to linear structure 2 (the angle is 0.88 degrees), linear structure 1 is basically perpendicular to linear structure 3 (the angle is 90.12 degrees), and linear structure 2 is basically perpendicular to linear structure 3 (the angle is 90.67 degrees). This is consistent with the structure of the space target, so the attitude of the space target can be determined.

[0063] Table 2 Numerical comparison of reconstruction results of two algorithms

[0064]

[0065] Obviously, those skilled in the art can make various changes and modifications to the present invention without departing from the spirit and scope of the present invention; thus, if these modifications and variations of the present invention fall within the scope of the claims of the present invention and their equivalents, the present invention is also intended to include these changes and variations.

Claims

1. A method for three-dimensional reconstruction of a target based on multi-view radar images, characterized in that: The following steps are involved: Step 1: The ISAR radar system receives echo signals within a set time period; performs range-Doppler processing on the echo signals within the set time period to obtain a radar image sequence of the target area, the sequence including P radar images of the target area; Step 2: Using a dual-parameter constant false alarm algorithm to distinguish target pixels from background pixels to obtain a binary segmentation image; using a morphological operator method to extract the target contour; Step 3, use parameter search method to perform centroid matching on ISAR images; Step 4, construct the target voxel using the spatial projection method; Specifically include: 4.1 Calculate the two-dimensional contour set G i The distance between them and the point Whether it exceeds the distance; if it exceeds the distance, the point is not considered to fall into the target area and output 0, otherwise go to step 4.2; Among them G i A set of two-dimensional contours consisting of edge points composition, is a point in the radar image; 4.2 Calculate the vector of each point j=1...J and 4.3 Determine the quadrant number Qu(j) of each edge point based on the point vector, which is 0 to 3; 4.4 Determine the vectors of two points and The symbol for the cross product is: The factor ε i =max{|r1|…|r j ||d1|…|d1|}×6.6613×10 -6 ; 4.5 Calculate the quadrant difference DQu(j) = Qu(j) - Qu(j+1) of each edge point and update it according to the following criteria: 4.6 Calculate the sum of the quadrant difference sequence. If the sum is not 0, the point is determined to be a point in the target area and the output is 1. Otherwise, the point is not in the target area and step 4.7 is executed. 4.7 Determining Point Vectors and The symbol of the inner product between If there exists j such that Sc(j)=0 and Inn(j)<0, the point is considered to be on the target contour and also a point in the target area, and the output is 1; otherwise, the output is 0.

2. The method for 3D target reconstruction based on multi-view radar images according to claim 1, wherein: Step 2 specifically includes: 2.1 Calculate the mean and variance of all pixels in the image; 2.2 Based on the calculated mean and variance, each pixel is divided into two categories: target and background according to the following formula; Threshold=μ+a1σ (1) Where Threshold is the threshold, μ is the mean value of pixels in the image, σ is the variance of pixels in the image, and a1 is the coefficient; 2.3 Based on the classification results, calculate the mean and variance of background pixels; 2.4 Based on the mean and variance in step 2.3, classify the background pixels again according to the following formula: Threshold'=μ′+a2σ′ (2) Where Threshold′ is the threshold, μ′ is the mean value of the pixels in the image in step 2.3, σ′ is the variance of the pixels in the image in step 2.3, and a2 is the coefficient; 2.5 Based on the classification results obtained in step 2.4, calculate the mean and variance of the background pixels again, obtain a new threshold and compare it with Threshold′. If the difference between the two is less than the given threshold, stop the iteration, otherwise return to step 2.3.

Images