A SAR target positioning and detection method based on radar and optical fusion

Abstract

The present invention discloses a SAR target positioning and detection method based on radar and optical fusion, which belongs to the field of target recognition technology, and includes the following steps: step 1, setting the center position of the radar and camera field angle on the same vertical line, controlling the radar and the camera to work at the same time, collecting radar data and optical data at the same time, and after the collection is completed, performing signal processing, radar imaging and imaging result enhancement on the collected radar signal to obtain a SAR image; performing data processing on the collected optical signal to obtain an optical image, and extracting optical target depth information at the same time; step 2, combining the data enhanced SAR image and the optical image for inter-frame alignment; step 3, after inter-frame alignment, converting the radar image target and the optical image target to the same spatial coordinate system to complete target positioning alignment; step 4, fusing the optical image target category to the SAR image to complete SAR target detection. The present invention improves the overall detection speed and accuracy.

Description

Technical Field

[0001] The invention belongs to the technical field of target recognition, and in particular relates to a SAR target positioning and detection method based on radar and optical fusion. Background Art

[0002] Millimeter-wave radar is an excellent sensor with high penetration and all-weather operation. It is widely used in the field of autonomous driving. By combining millimeter-wave radar with synthetic aperture radar technology, it is installed on platforms such as cars and drones to achieve high-resolution imaging of near-field SAR, calculate the distance, speed, angle and other information of the target, and achieve all-round perception of the road environment. Radar perceives the surrounding environment by emitting electromagnetic waves and receiving reflected signals. Therefore, its viewing angle is usually wide and can detect objects at different distances and directions, but the detection of detailed information such as the shape and texture of the object is not fine enough. Optical cameras have a simple structure and mature design. They can provide clear and detailed images and can obtain target feature details more intuitively. It is of certain significance to analyze the geometric and scale differences between radar images and optical images and integrate the basis of optical and radar target fusion into SAR target detection for research.

[0003] In the paper "Ship Detection Based on Polarimetric SAR Gradient and Complex Wishart Classifier", Yin Junjun et al. proposed to introduce the likelihood ratio test (LRT) gradient into the logarithmic ratio gradient framework and perform constant false alarm (CFAR) detection based on the LRT gradient map to extract the edge information of the ship, eliminate artifacts, and suppress the influence of strong side lobes on the extraction of fine ship contours; in addition, the strong scattering part of the ship is detected using a complex Wishart iterative classifier to eliminate most of the clutter interference and maintain the details of the ship's shape, which can have fewer false alarms and missed detections. However, for near-field SAR targets, due to the complex structure of each target and the existence of multiple scattering mechanisms, the imaging results change with the scattering conditions such as the incident angle and radar system parameters. Therefore, the scattering results of the same target in different frames may also be very different. New solutions are still needed to solve the problem of false detection and missed detection of near-field SAR targets.

[0004] Saurabh Saxena et al. explored the use of high-fidelity denoising diffusion models in monocular depth estimation in the paper "Monocular Depth Estimation using Diffusion Models", and used the image sampler of the spatial transformation network to reconstruct the input image in three dimensions, achieving certain results in monocular depth estimation. However, the network results designed by the researchers were ultimately presented in the form of image RGB values, and the actual depth data was not obtained. Further research is still needed on the depth distance information required for radar and optical coordinate mapping.

[0005] In the paper “Microwave and Optical Interactive Perception and Recognition System for Intelligent Driving”, Zhang Lei built a micro-vehicle platform, proposed edge radar real-time imaging processing and edge camera deep target recognition processing, carried out microwave and optical information fusion, and established an information interaction channel between radar and camera. However, the information fusion method proposed by the researcher only stopped at the conversion of the center point of radar and optical space coordinates, and did not combine the radar target morphology for analysis and research. Summary of the invention

[0006] The purpose of the present invention is to provide a SAR target positioning and detection method based on radar and optical fusion, so as to solve the problems in the prior art of near-field SAR target detection based on radar and optical fusion, such as high hardware cost, slow speed, difficulty in obtaining actual depth data and failure to analyze and study the radar target morphology.

[0007] To achieve the above object, the present invention provides a SAR target positioning and detection method based on radar and optical fusion, comprising the following steps:

[0008] Step 1: Set the center positions of the radar and camera field of view angles on the same vertical line, control the radar and camera systems to work simultaneously, collect radar data and optical data simultaneously, and after the collection is completed, perform signal processing, radar imaging and imaging result enhancement on the collected radar signals to obtain SAR images; perform data processing on the collected optical signals to obtain optical images, and extract the depth information of optical targets at the same time;

[0009] Step 2: align the SAR image and the optical image after data enhancement;

[0010] Step 3: After inter-frame alignment, the radar image target and the optical image target are converted to the same spatial coordinate system to complete target positioning alignment;

[0011] Step 4: Fuse the optical image target category into the SAR image to complete SAR target detection.

[0012] Preferably, the specific process of obtaining the SAR image in step 1 is as follows:

[0013] SA1, the radar transmits a broadband signal to the detection area. The signal is received by the radar after being reflected by the target, and high-resolution range imaging is performed through the synthetic aperture radar SAR algorithm. At this time, the SAR image solves the target range and azimuth information, which corresponds to the optical depth dimension and distance dimension information;

[0014] SA2, save the matrix after imaging as the original data matrix or perform nonlinear transformation on the data matrix after imaging; analyze the original morphological characteristics of the SAR target according to the original data matrix to perform target detection; complete data enhancement on the data matrix after nonlinear transformation processing as the basis for inter-frame alignment;

[0015] SA3. Extract the original data matrix of the SAR image and the data matrix after data enhancement into single-frame SAR images and save them for later use.

[0016] Preferably, in step 1, optical data is collected by a monocular camera or a binocular camera, and the specific process of obtaining an optical image by performing data processing on the optical signal collected by the monocular camera is as follows:

[0017] SB1. The camera collects 2D image data of the actual scene, corrects image distortion based on the camera matrix parameters, and passes the corrected image to the depth estimation network to complete 3D reconstruction and obtain depth dimension data.

[0018] SB2, the depth estimation network obtains the depth map image of the scene, uses the photometric reconstruction error loss function to control the difference between the target image and the reconstructed target image, and optimizes the depth estimation model in combination with the actual scene parameters;

[0019] SB3, through the data shift algorithm, the depth map image information is converted into actual depth data to complete the extraction of depth information;

[0020] SB4. Use the YOLOv5 target detection network to map the target information identified in the two-dimensional image to the depth image, and use the method of proportionally reducing the target detection box on the depth image to obtain the depth information of the specific target on the depth image.

[0021] Preferably, in SB4, the calculation expression for obtaining the depth information of a specific target on the depth image by scaling down the target detection frame on the depth image is as follows:

[0022]

[0023] The set S represents all pixel values ​​inside the target frame in the depth map whose pixel values ​​are between the maximum and minimum values, a is used to adjust the maximum error, and ∑x represents the total number of all pixel values ​​that meet the conditions.

[0024] Preferably, the specific expression of the photometric reconstruction error loss function in SB2 is as follows:

[0025]

[0026] Among them, L p is the photometric reconstruction error loss function, I t is the target image, I t'is the source image.

[0027] Preferably, when the binocular camera collects optical data, it uses a camera imaging model algorithm, binocular data stereo correction and three-dimensional reconstruction algorithm, combined with an optical YOLOv5 network to detect target position and target category information, and finally obtains the depth distance information of the designated target center point.

[0028] Preferably, the specific process of inter-frame alignment in step 2 is as follows:

[0029] S21, finding the start frame and the end frame of the actual scene, and determining the optical images corresponding to the first frame and the last frame of the SAR image using the corresponding relationship between the center target of the field of view angle;

[0030] S22, assuming that the field of view of the optical image is x, and the field of view of the SAR image is set to y, satisfying y>x. For each frame of imaging viewing angle, the SAR image scene covers targets including the optical image scene targets. After confirming the starting frame, the inter-frame alignment algorithm is used to filter the images between the first frame and the last frame in the actual scene, so that each frame of the SAR image has a corresponding optical image. The specific expression of the inter-frame alignment algorithm is as follows:

[0031]

[0032] Among them, N r Indicates the SAR image label, S r is the total number of frames of SAR images in the actual process, t i is the acquisition time in the optical image, Δt is the acquisition time of the alignment of optical and radar data acquisition times, fps i is the frame rate of the camera capturing images per second, S i The total number of frames of optical images in the actual scene;

[0033] S23. After inter-frame correction processing, the SAR image and the optical image of the actual scene maintain a center alignment standard in each frame, and a synchronization relationship is established between the radar scene and the optical scene.

[0034] Preferably, in step 3, the specific process of converting the radar image target and the optical image target to the same spatial coordinate system to complete the target positioning alignment is as follows:

[0035] S31, calculating the angle of the target in the center of the optical target detection frame deviating from the center of the line of sight, and calculating the proportion of the size of the frame image occupied by the pixel coordinate point of the target in the center of the optical scene, and obtaining the angle ratio Δθ of the target deviating from the center of the line of sight on the SAR image. The specific expression of the angle ratio is as follows:

[0036]

[0037] Where η is the camera field of view, x t is the horizontal size of the frame image, x0 is the horizontal coordinate of the pixel coordinate point of the target at the center of the optical scene;

[0038] S32. By using the quantitative relationship between the camera field of view angle and the SAR image field of view angle, and combining the fact that the distances of the same target from the midpoint of the horizontal axis of the image pixel distribution in the azimuth direction in the SAR image and the optical image are consistent, the vertical axis pixel value is solved to confirm the location of the target on the SAR image; the expression of the converted SAR image target point (x, y) is as follows:

[0039]

[0040] Among them, x p ,y p is the pixel coordinate of the horizontal and vertical axes of the target in the optical image, r radar is the actual maximum working distance of SAR image, x t is the horizontal size of the frame image, y t is the vertical size of the frame image, r p is the pixel distance of the farthest effective distance on the SAR image.

[0041] Preferably, the specific process of fusing the optical image target information with the SAR image target information in step 4 is as follows:

[0042] S41, after completing the positioning alignment of the optical image target and the radar image target, find the target location on the radar image with the guidance of the optical positioning, and complete the positioning fusion of the optical image target and the radar image target;

[0043] S42. After positioning and fusion, target category information obtained by optical image target detection is annotated into SAR image target positioning information to complete the fusion of optical image target and radar image target information; SAR target detection is completed with the guidance of optical image target positioning and category information.

[0044] Therefore, the present invention adopts the above-mentioned SAR target positioning and detection method based on radar and optical fusion, which has the following beneficial effects:

[0045] (1) Taking advantage of the deep neural network's autonomous learning of image features in depth estimation, it can overcome interference from factors such as lighting and texture to a certain extent and obtain relatively accurate depth information;

[0046] (2) Since the angles of the same target in the SAR image frame and the optical image frame relative to the central coordinates are proportional, the proportional relationship in different coordinate systems is used to establish equations, and the depth dimension information and azimuth dimension information of the optical image and the range and azimuth dimensions of the SAR image are converted to the same coordinate system for analysis. The proportional relationship conversion of spatial targets has a certain improvement in speed compared to the previous approximate evaluation method of infinite division of the azimuth distance.

[0047] The technical solution of the present invention is further described in detail below through the accompanying drawings and embodiments. BRIEF DESCRIPTION OF THE DRAWINGS

[0048] Figure 1 It is an overall flow chart of a SAR target positioning and detection method based on radar and optical fusion of the present invention;

[0049] Figure 2 It is a depth estimation effect diagram of the present invention;

[0050] Figure 3 It is a comparison diagram of the optical imaging scene and the radar imaging scene of the present invention;

[0051] Figure 4 This is the automobile target fusion effect diagram of the present invention. DETAILED DESCRIPTION

[0052] The following detailed description of the embodiments of the present invention provided in the accompanying drawings is not intended to limit the scope of the invention claimed for protection, but merely represents selected embodiments of the present invention. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0053] See also Figure 1-4 , a SAR target positioning and detection method based on radar and optical fusion, comprising the following steps:

[0054] Step 1: Set the center positions of the radar and camera field of view angles on the same vertical line, control the radar and camera systems to work simultaneously, collect radar data and optical data simultaneously, and after the collection is completed, perform signal processing, radar imaging and imaging result enhancement on the collected radar signals to obtain SAR images; perform data processing on the collected optical signals to obtain optical images, and extract the depth information of optical targets at the same time;

[0055] The specific process of obtaining SAR images is as follows:

[0056] SA1, the radar transmits a broadband signal to the detection area. The signal is received by the radar after being reflected by the target, and high-resolution range imaging is performed through the synthetic aperture radar SAR algorithm. At this time, the SAR image solves the target range and azimuth information, which corresponds to the optical depth dimension and distance dimension information;

[0057] SA2, save the matrix after imaging as the original data matrix or perform nonlinear transformation on the data matrix after imaging; analyze the original morphological characteristics of the SAR target according to the original data matrix to perform target detection; complete data enhancement on the data matrix after nonlinear transformation processing as the basis for inter-frame alignment;

[0058] SA3. Extract the original data matrix of the SAR image and the data matrix after data enhancement into single-frame SAR images and save them for later use.

[0059] The specific process of collecting optical data through a monocular camera or a binocular camera and processing the optical signal collected by the monocular camera to obtain an optical image is as follows:

[0060] SB1. The camera collects 2D image data of the actual scene, corrects image distortion based on the camera matrix parameters, and passes the corrected image to the depth estimation network to complete 3D reconstruction and obtain depth dimension data.

[0061] SB2. The depth estimation network obtains the depth map image of the scene, controls the difference between the target image and the reconstructed target image by controlling the photometric reconstruction error loss function, and optimizes the depth estimation model in combination with the actual scene parameters; the specific expression of the photometric reconstruction error loss function is as follows:

[0062]

[0063] Among them, L p is the photometric reconstruction error loss function, I t is the target image, I t' is the source image;

[0064] SB3, through the data shift algorithm, the depth map image information is converted into actual depth data to complete the extraction of depth information;

[0065] SB4. Use the YOLOv5 target detection network to map the target information identified on the two-dimensional image to the depth image, and use the method of proportionally reducing the target detection frame on the depth image to obtain the depth information of the specific target on the depth image. The specific calculation expression is as follows:

[0066]

[0067] The set S represents all pixel values ​​between the maximum and minimum values ​​inside the target frame in the depth map, a is used to adjust the maximum error, and ∑x represents the total number of all pixel values ​​that meet the conditions. The depth estimated pixel values ​​of different scenes are fitted according to the actual depth distance and converted into the actual values ​​of the scene, such as Figure 2 Shown is the depth estimation effect diagram.

[0068] When the binocular camera collects optical data, it uses the camera imaging model algorithm, binocular data stereo correction and 3D reconstruction algorithm, combined with the optical YOLOv5 network to detect the target position and target category information, and finally obtains the depth distance information of the specified target center point.

[0069] Step 2: Combine the data-enhanced SAR image and the optical image to perform frame alignment. The specific process is as follows:

[0070] S21, finding the start frame and the end frame of the actual scene, and determining the optical images corresponding to the first frame and the last frame of the SAR image using the corresponding relationship between the center target of the field of view angle;

[0071] S22, assuming that the field of view of the optical image is x, and the field of view of the SAR image is set to y, satisfying y>x. For each frame of imaging viewing angle, the SAR image scene covers targets including the optical image scene targets. After confirming the starting frame, the inter-frame alignment algorithm is used to filter the images between the first frame and the last frame in the actual scene, so that each frame of the SAR image has a corresponding optical image. The specific expression of the inter-frame alignment algorithm is as follows:

[0072]

[0073] Among them, N r Indicates the SAR image label, S r is the total number of frames of SAR images in the actual process, t i is the acquisition time in the optical image, Δt is the acquisition time of the alignment of optical and radar data acquisition times, fps i is the frame rate of the camera capturing images per second, S i The total number of frames of optical images in the actual scene;

[0074] S23. After inter-frame correction processing, the SAR image and the optical image of the actual scene maintain a center alignment standard in each frame, and a synchronization relationship is established between the radar scene and the optical scene.

[0075] Step 3: After inter-frame alignment, use the YOLOv5 algorithm to identify cars, bicycles, and pedestrians in the optical image, and extract the depth distance information of the target in the depth data matrix based on the target prior frame position information. At this time, the optical target has obtained the azimuth and depth dimension data information. In order to mark the accurate position of the optical target on the SAR image so that the optical category can guide the SAR image target detection, the radar image target and the optical image target need to be converted to the same spatial coordinate system to complete the target positioning alignment; the specific process is as follows:

[0076] S31, calculating the angle of the target in the center of the optical target detection frame deviating from the center of the line of sight, and calculating the proportion of the size of the frame image occupied by the pixel coordinate point of the target in the center of the optical scene, and obtaining the angle ratio Δθ of the target deviating from the center of the line of sight on the SAR image. The specific expression of the angle ratio is as follows:

[0077]

[0078] Where η is the camera field of view, x t is the horizontal size of the frame image, x0 is the horizontal coordinate of the pixel coordinate point of the target at the center of the optical scene;

[0079] S32, by using the quantitative relationship between the camera field of view angle and the SAR image field of view angle, combined with the consistent distance of the same target from the midpoint of the horizontal axis of the image pixel distribution in the SAR image and the optical image, solving the vertical axis pixel value to confirm the location of the target on the SAR image; Figure 3 As shown, the expression of the converted SAR image target point (x, y) is as follows:

[0080]

[0081] Among them, x p ,y p is the pixel coordinate of the horizontal and vertical axes of the target in the optical image, r radar is the actual maximum working distance of SAR image, x t is the horizontal size of the frame image, y t is the vertical size of the frame image, r p is the pixel distance of the farthest effective distance on the SAR image. According to the running time comparison of different algorithms for the same scene, the running time of the distance-azimuth relationship mapping algorithm is 146.25 seconds, and the running time of the angle coordinate mapping algorithm is 38.34 seconds. The use of the angle coordinate mapping algorithm is 73.97% faster than the distance-azimuth relationship mapping algorithm.

[0082] Step 4: Fuse the optical image target category into the SAR image to complete SAR target detection, such as Figure 4 As shown, the specific process is as follows:

[0083] S41, after completing the positioning alignment of the optical image target and the radar image target, find the target location on the radar image with the guidance of the optical positioning, and complete the positioning fusion of the optical image target and the radar image target;

[0084] S42. After positioning and fusion, target category information obtained by optical image target detection is annotated into SAR image target positioning information to complete the fusion of optical image target and radar image target information; SAR target detection is completed with the guidance of optical image target positioning and category information.

[0085] Therefore, the present invention adopts the above-mentioned SAR target positioning and detection method based on radar and optical fusion, and establishes equations by depth estimation and using proportional relationships in different coordinate systems. The depth dimension information and azimuth dimension information of the optical image and the distance and azimuth dimensions of the SAR image are converted to the same coordinate system for analysis, and the proportional relationship conversion of space targets is used to improve the overall detection speed and accuracy.

[0086] Finally, it should be noted that the above embodiments are only used to illustrate the technical solution of the present invention rather than to limit it. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that they can still modify or replace the technical solution of the present invention with equivalents, and these modifications or equivalent replacements cannot cause the modified technical solution to deviate from the spirit and scope of the technical solution of the present invention.

Claims

1. A SAR target positioning and detection method based on radar and optical fusion, characterized in that: The following steps are involved: Step 1: Set the center positions of the radar and camera field of view angles on the same vertical line, control the radar and camera systems to work simultaneously, collect radar data and optical data simultaneously, and after the collection is completed, perform signal processing, radar imaging and imaging result enhancement on the collected radar signals to obtain SAR images; Perform data processing on the collected optical signals to obtain optical images and extract the depth information of the optical target; Step 2: align the SAR image and the optical image after data enhancement; Step 3: After inter-frame alignment, the radar image target and the optical image target are converted to the same spatial coordinate system to complete target positioning alignment; Step 4: Fuse the optical image target category into the SAR image to complete SAR target detection; The specific process of inter-frame alignment in step 2 is as follows: S21, finding the start frame and the end frame of the actual scene, and determining the optical images corresponding to the first frame and the last frame of the SAR image using the corresponding relationship between the center target of the field of view angle; S22, let the field of view of the optical image be x 视 , the SAR image field of view is set to y 视 , satisfying y 视 >x 视 , for each frame of imaging perspective, the SAR image scene covers targets including optical image scene targets; after confirming the starting frame, the inter-frame alignment algorithm is used to filter the images between the first frame and the last frame in the actual scene, so that each frame of SAR image has a corresponding optical image. The specific expression of the inter-frame alignment algorithm is as follows: ； in, Indicates the SAR image number, is the total number of frames of SAR images in the actual process, is the acquisition moment in the optical image, is the acquisition time when optical and radar data are collected in real time. is the frame rate at which the camera collects images per second, The total number of frames of optical images in the actual scene; S23, after inter-frame correction processing, the SAR image and the optical image of the actual scene maintain the center alignment standard in each frame, and the radar scene and the optical scene establish a synchronization relationship; The specific process of converting the radar image target and the optical image target to the same spatial coordinate system to complete the target positioning alignment in step 3 is as follows: S31, calculate the angle of the target in the center of the optical target detection frame that deviates from the center of the line of sight, calculate the proportion of the coordinate point in the frame image size occupied by the pixel coordinate point of the target in the center of the optical scene, and obtain the angle ratio of the target in the SAR image that deviates from the center of the line of sight ; S32. By using the quantitative relationship between the camera field of view angle and the SAR image field of view angle and combining the fact that the distance of the same target from the midpoint of the horizontal axis of the image pixel distribution in the azimuth direction in the SAR image and the optical image is consistent, the vertical axis pixel value is solved to confirm the location of the target on the SAR image.

2. The SAR target positioning and detection method based on radar and optical fusion according to claim 1, characterized in that: The specific process of obtaining the SAR image in step 1 is as follows: SA1, the radar transmits a broadband signal to the detection area. The signal is received by the radar after being reflected by the target, and high-resolution range imaging is performed through the synthetic aperture radar SAR algorithm. At this time, the SAR image solves the target range and azimuth information, which corresponds to the optical depth dimension and distance dimension information; SA2, saving the imaged matrix as the original data matrix or performing nonlinear transformation on the imaged data matrix; The original data is used to analyze the original morphological features of the SAR target and perform target detection; The data matrix after nonlinear transformation completes data enhancement as the basis for inter-frame alignment; SA3. Extract the original data matrix of the SAR image and the data matrix after data enhancement into single-frame SAR images and save them for later use.

3. The SAR target positioning and detection method based on radar and optical fusion according to claim 2 is characterized in that: In step 1, optical data is collected by a monocular camera or a binocular camera, and the specific process of obtaining an optical image by processing the optical signal collected by the monocular camera is as follows: SB1. The camera collects 2D image data of the actual scene, corrects image distortion based on the camera matrix parameters, and passes the corrected image to the depth estimation network to complete 3D reconstruction and obtain depth dimension data. SB2, the depth estimation network obtains the depth map image of the scene, controls the difference between the target image and the reconstructed target image by controlling the photometric reconstruction error loss function, and optimizes the depth estimation model in combination with the actual scene parameters; SB3, through the data shift algorithm, the depth map image information is converted into actual depth data to complete the extraction of depth information; SB4. Use the YOLOv5 target detection network to map the target information identified in the two-dimensional image to the depth image, and use the method of proportionally reducing the target detection box on the depth image to obtain the depth information of the specific target on the depth image.

4. The SAR target positioning and detection method based on radar and optical fusion according to claim 3 is characterized in that: In SB4, the calculation expression for obtaining the depth information of the specific target on the depth image by reducing the target detection frame in equal proportion on the depth image is as follows: ； The collection Indicates all pixel values ​​inside the target frame in the depth map that are between the maximum and minimum values. Used to adjust the maximum error. Represents the total number of pixel values ​​that meet the conditions. Represents the pixel value of each pixel.

5. The SAR target positioning and detection method based on radar and optical fusion according to claim 4, characterized in that: The specific expression of the photometric reconstruction error loss function in SB2 is as follows: ； in, is the photometric reconstruction error loss function, is the target image, is the source image.

6. The SAR target positioning and detection method based on radar and optical fusion according to claim 5, characterized in that: When the binocular camera collects optical data, it uses the camera imaging model algorithm, binocular data stereo correction and 3D reconstruction algorithm, combined with the optical YOLOv5 network to detect the target position and target category information, and finally obtains the depth distance information of the specified target center point.

7. The SAR target positioning and detection method based on radar and optical fusion according to claim 6, characterized in that: The specific process of fusing the optical image target information with the SAR image target information in step 4 is as follows: S41, after completing the positioning alignment of the optical image target and the radar image target, find the target location on the radar image with the guidance of the optical positioning, and complete the positioning fusion of the optical image target and the radar image target; S42. After positioning and fusion, target category information obtained by optical image target detection is annotated into SAR image target positioning information to complete the fusion of optical image target and radar image target information; SAR target detection is completed with the guidance of optical image target positioning and category information.

Images