Unmanned collaborative detection method and system based on SAR-guided optical payload

Abstract

The present invention discloses an unmanned collaborative detection method and system based on SAR-guided optical payloads, comprising the following steps: S1: acquiring and demodulating echo signals through a SAR antenna; S2: inputting the demodulated echo signals into an FPGA to obtain a SAR image; S3: transmitting the SAR image to an edge computing platform; S4: performing target detection based on YOLO-TensorRT; determining whether a target is detected; S5: determining the final target through confidence screening and circumscribed circle geometry verification; S6: performing pixel coordinate to geographic coordinate conversion to obtain the WGS-84 geographic coordinates of the final target; S7: the unmanned aerial vehicle receives an approach flight path instruction from a ground station and flies toward the target area; the optical payload activates a high-definition imaging mode to verify target details. This invention can solve technical problems such as poor real-time performance, low resource coordination efficiency, and inaccurate target guidance in multi-sensor collaborative detection in complex scenarios.

Description

Technical Field

[0001] The present invention relates to the fields of unmanned aerial vehicle (UAV) detection and synthetic aperture radar (SAR) image processing, and in particular to an unmanned collaborative detection method and system based on SAR-guided optical payloads. Background Art

[0002] The rapid development of unmanned technology has driven its extensive application in strategic areas such as military reconnaissance, disaster monitoring, and border patrol. However, target detection in complex scenarios faces multi-faceted technical challenges. From an environmental perspective, electromagnetic multipath effects from urban buildings, obstruction from dense vegetation, and the varying elevations of mountainous terrain make it difficult for a single sensor to meet practical detection requirements. In this technological context, multi-sensor collaborative detection has become a key path to overcoming performance bottlenecks. Synthetic aperture radar (SAR) and optical payload collaboration offer unique advantages. SAR is an active microwave remote sensing technology that can acquire large-area SAR imagery in all weather conditions and at all times. Recent advances in micro-SAR technology have enabled wide-area real-time imaging. Optical payload pods, typically those using visible / infrared light, provide high-resolution imaging and accurate target identification, but their viewing angles are narrow and resolution is insufficient. Collaborative approaches based on SAR-guided optical payloads theoretically offer the potential to balance wide-area search efficiency with target recognition accuracy. However, practical implementation still faces key technical bottlenecks, such as hardware collaboration and real-time processing.

[0003] First, there are problems with real-time and resource coordination efficiency in the collaborative detection of SAR and optical payloads. SAR imaging requires processing massive amounts of raw data, and if transmitted back, it will compete with flight control instructions within a limited bandwidth. Traditional imaging, positioning, and recognition algorithms are all deployed on the same hardware platform, which can easily lead to resource overload and fluctuating processing performance under high temperatures. Therefore, how to design heterogeneous computing architectures and dynamic resource scheduling to achieve efficient collaboration has become the key to breaking through real-time bottlenecks and ensuring stable system operation. Reviewing existing related patent technologies, CN119289984A proposed a multi-payload task planning optimization algorithm, but its core limitation is that it has not built an efficient hardware coordination architecture. It only coordinates sensor work through traditional path planning, which cannot solve the problems of real-time processing and resource competition. CN119339067A focuses on SAR-photoelectric collaboration, but it does not design a dedicated hardware coordination architecture and relies on SAR post-processing images. Although the current patented technologies for SAR and optical payload data fusion (CN119762557A proposes visible light and SAR registration, and CN119762558A proposes infrared and SAR registration) have built a registration and fusion algorithm framework based on multi-scale descriptors, they have not built a hardware collaborative architecture and have limitations in engineering adaptability.

[0004] It can be seen from this that although the existing patented technology has realized the mission planning and data fusion of SAR and optical payloads, it only focuses on the data processing flow at the algorithm level, and does not fully consider the hardware coordination mechanism, dynamic environment adaptation and engineering implementation requirements in the actual application of UAVs. There are problems such as insufficient real-time performance, inefficient resource scheduling, and poor robustness in complex scenarios. Summary of the Invention

[0005] In response to the above, the present invention proposes an unmanned collaborative detection method and system based on SAR-guided optical payload. Through a full-process real-time processing chain, a closed loop from SAR raw echo acquisition to target geographic coordinate guidance is realized, solving technical problems such as poor real-time performance, low resource coordination efficiency, and inaccurate target guidance in multi-sensor collaborative detection in complex scenarios.

[0006] The present invention proposes an unmanned collaborative detection method based on SAR-guided optical payload, comprising the following steps:

[0007] S1: The SAR antenna is used to scan the target area in strips to obtain echo signals, and the echo signals are demodulated to obtain demodulated echo signals;

[0008] S2: Input the demodulated echo signal into FPGA, perform range and azimuth compression to realize imaging, and obtain SAR image;

[0009] S3: Transmit the SAR image to the edge computing platform via the COFDM data transmission link of the data communication radio;

[0010] S4: Perform INT8 quantization on the lightweight YOLOv11 model to obtain YOLO-TensorRT; perform target detection on the SAR image based on YOLO-TensorRT; determine whether a target is detected: if a target exists, proceed to S5; if no target is detected, return to S1;

[0011] S5: Confidence screening of target detection results and geometric verification of the circumscribed circle to determine the final target;

[0012] S6: Combined with the SAR image, the coordinates of the center of the circumscribed circle of the final target are used as the input for pixel coordinate to geographic coordinate conversion, and the pixel coordinate to geographic coordinate conversion is performed, and the elevation correction is performed using the digital elevation model data to obtain the WGS-84 geographic coordinates of the final target;

[0013] S7: The edge computing platform transmits the WGS-84 geographic coordinates of the final target to the ground station through the ground base station. The UAV receives the approach flight path instructions forwarded by the ground station through the ground base station, adjusts the heading and altitude, and flies towards the target area; the optical payload starts the high-definition imaging mode to verify the target details.

[0014] Furthermore, the S1 includes:

[0015] S11: The radar signal is transmitted to the target area through the SAR antenna. The echo signal reflected by the target and received is expressed as follows:

[0016] ;

[0017] in, It represents the echo signal received by SAR after being reflected by the target; is the target reflection coefficient, is a complex constant, t, Represent the slow time dimension and the fast time dimension respectively, is the moment when the center of the beam passes the target, and c is the propagation speed of the electromagnetic wave; Indicates the azimuth signal reception strength, where a represents the azimuth, which is the direction related to the SAR movement direction; Indicates the signal distance pulse envelope strength; is the slant range of SAR to the target at time t; exp() is the exponential function; is the center frequency, is the range modulation frequency; and They represent the signal range pulse envelope function and the azimuth signal receiving strength function respectively, and their expressions are as follows:

[0018] ;

[0019] in, is the pulse duration, P a is the azimuth antenna pattern function, rect() is the rectangular function, Indicates when When; sinc is the Sinker function; is the angle between the target and the beam centerline at time t; is the azimuth beamwidth, is the signal wavelength, is the antenna width in azimuth;

[0020] S12: Demodulate the echo signal to obtain a demodulated echo signal:

[0021] ;

[0022] in, represents the demodulated echo signal; j is the imaginary unit.

[0023] Furthermore, the S2 includes:

[0024] The demodulated echo signal is input into the FPGA as the SAR raw data, and the range and azimuth compression are performed to realize imaging. The specific steps are as follows:

[0025] S21: distance compression;

[0026] Perform FFT on the distance direction of the demodulated echo signal and obtain the following expression:

[0027] ;

[0028] in, is the distance of the demodulated echo signal and the echo signal after FFT processing; G is the total gain including the scattering coefficient, which is set to 1; is the envelope of the range spectrum, is the distance frequency; is the slant range between the SAR and the target at the slow time η;

[0029] Range-matched filter The expression is:

[0030] ;

[0031] After FFT processing, the echo signal is output through the range filter and then through IFFT to obtain the range compressed expression. :

[0032] ;

[0033] in, is the inverse fast Fourier transform; for IFFT expression of ;

[0034] S22: azimuth FFT;

[0035] SAR slant range to target :

[0036] ;

[0037] in, is the initial slant range, v is the SAR platform speed;

[0038] Since the beam is pointing close to the zero Doppler direction, and ,Will Approximately a parabola, the slant range of the SAR from the target , the calculation formula can be replaced by:

[0039] ;

[0040] When performing azimuth FFT, the time-frequency relationship is: ,in, is the azimuth frequency; is the frequency modulation in azimuth direction; is the movement speed of the SAR platform; is the wavelength of the SAR transmitted signal; the azimuth FFT expression is as follows:

[0041] ;

[0042] in, is the echo signal after FFT processing in azimuth; To perform fast Fourier transform on the time variable t; is the echo signal after range compression; is the azimuth frequency variable; for Frequency domain form of This is the distance migration compensation term, and its expression is as follows:

[0043] ;

[0044] S23: Perform distance migration correction;

[0045] The amount of distance migration that needs to be corrected , the expression is as follows:

[0046] ;

[0047] Range migration correction factor , the expression is as follows:

[0048] ;

[0049] The echo signal after range migration correction is , the expression is as follows:

[0050] ;

[0051] S24: Azimuthal compression.

[0052] Furthermore, the S24 azimuth compression includes:

[0053] Azimuth matched filter The design is as follows:

[0054] ;

[0055] The echo signal after the azimuth matched filter output for:

[0056] ;

[0057] Finally, the final compressed signal is obtained through azimuth IFFT for:

[0058] ;

[0059] in, is the amplitude of the azimuthal impulse response;

[0060] After range compression and azimuth compression imaging processing, the final target has been corrected to At this point, target imaging processing is completed and a SAR image is obtained.

[0061] Furthermore, the S5 includes:

[0062] After target detection based on YOLO-TensorRT, the model outputs a sequence of predicted boxes , where B is the prediction box sequence, is the i-th prediction box, and m is the total number of prediction boxes; ,in and is the coordinate of the center point of the i-th prediction box, and are the width and height of the i-th prediction box, is the confidence of the i-th prediction box, is the target category probability;

[0063] Through confidence screening and geometric verification of the circumscribed circle, abnormal targets are eliminated and the final target is determined. Specifically:

[0064] S51: confidence screening;

[0065] Set a confidence threshold , which is used to filter out prediction frames with higher confidence and obtain a valid prediction frame set; the filtering conditions are as follows:

[0066] ;

[0067] in, It is the set of valid prediction boxes obtained after confidence screening;

[0068] S52: Circumscribed circle calculation and verification;

[0069] S521: Calculation of circumscribed circle;

[0070] For the valid prediction box set Each valid prediction box in , calculate the coordinates of the center of its circumscribed circle and radius ; Since the effective prediction box is a rectangle, the center of its circumscribed circle is the center point of the rectangle, and the radius is the distance from the center point to the vertex of the rectangle, specifically:

[0071] The formula for calculating the coordinates of the circle center is:

[0072] ;

[0073] Radius calculation formula:

[0074] Assume that a vertex of the valid prediction box is , according to the distance formula between two points, the radius for:

[0075] ;

[0076] S522: Geometric verification of circumscribed circle;

[0077] Detect the same target in N consecutive SAR images. Let the coordinates of the center of the circumscribed circle of the nth frame be , the radius is ; Calculate the mean of the center coordinates of N consecutive frames and the mean of the radius :

[0078] ;

[0079] Setting distance threshold and radius change threshold ; For each frame's center coordinates and radius , and calculate its deviation from the mean:

[0080] Center coordinate deviation :

[0081] ;

[0082] Radius deviation :

[0083] ;

[0084] like or , then the detection result of the frame is considered to be abnormal, and the valid prediction frame corresponding to the SAR image of the frame is removed;

[0085] S53: Final goal determined;

[0086] After confidence screening and circumscribed circle geometry verification, the target corresponding to the retained valid prediction box is the final target.

[0087] Furthermore, the S6 includes:

[0088] S61: Geocoding model;

[0089] Assume that the longitude, latitude and altitude measured by GNSS of the UAV platform are , the error correction values ​​of longitude, latitude and altitude calculated by the ground base station are , then the position of the UAV platform after differential correction is:

[0090] ;

[0091] in, , , are the longitude, latitude and altitude of the platform after differential correction;

[0092] Combined with the SAR image, the coordinates of the center of the circumscribed circle of the final target are As input to the pixel coordinate to geographic coordinate conversion, establish pixel coordinates Mapping relationship to geographic space;

[0093] Assuming the platform position is the origin, calculate the initial latitude and longitude of the target using the spherical trigonometry formula :

[0094] ;

[0095] in, and are the initial longitude increment and latitude increment respectively; R e is the average radius of the earth; R is the slant range of the SAR to the target; φ is the azimuth of the target relative to the platform, according to the pixel coordinates Determination of the position in the SAR image and the geometric relationship of SAR imaging; is the cosine value of the UAV platform latitude after differential correction; is the beam pitch angle in SAR imaging; sin and cos are sine and cosine functions respectively;

[0096] The initial latitude and longitude coordinates are:

[0097] ;

[0098] S62: elevation correction;

[0099] The SRTM 30m resolution digital elevation model data was integrated to correct the terrain relief of the preliminary latitude and longitude coordinates. Specifically:

[0100] Assume that the target area elevation value obtained from SRTM data is , the coordinate deviations caused by the terrain are and ;

[0101] According to the geometric relationship, the coordinate deviation can be approximately calculated:

[0102] ;

[0103] Among them, tan is the tangent function;

[0104] Get the WGS-84 geographic coordinates (lon, lat) of the final target:

[0105] .

[0106] Furthermore, the S7 includes:

[0107] S71: The edge computing platform transmits the WGS-84 geographic coordinates of the final target to the ground station via the ground base station. After receiving the WGS-84 geographic coordinates of the final target, the ground station generates the approach flight path of the UAV using a path planning algorithm.

[0108] S72: The UAV receives the approach flight path command forwarded by the ground station via the ground base station, adjusts its heading and altitude, and flies toward the target area. The optical payload activates high-definition imaging mode to verify target details.

[0109] Ground base stations monitor the signal strength of communication links in real time and bit error rate BER;

[0110] Let the signal strength threshold be , the bit error rate threshold is ;

[0111] when or When the signal is transmitted, the ground base station dynamically switches channels or enhances signal power.

[0112] The present invention also discloses an unmanned collaborative detection system based on SAR-guided optical payload, the system comprising: a drone, a SAR, a data communication radio, a voltage converter, an edge computing platform, a micro POS system, a power module, a radio frequency module and an optical payload;

[0113] The SAR is installed on a drone and includes a SAR antenna and a SAR host. The SAR antenna serves as the front-end component of the SAR and is used to transmit and receive radar signals. The SAR host is composed of an FPGA motherboard, provides a Gigabit Ethernet connection to the edge computing platform, integrates a radar waveform generation module, and supports dynamic adjustment of the pulse repetition frequency. The SAR obtains a SAR image after the signal received by the SAR antenna 1 is processed by the FPGA of the SAR host.

[0114] The digital communication station is provided with a digital communication station antenna for realizing data transmission between the SAR and the ground station;

[0115] The voltage converter is located below the digital communication station, and is connected to the digital communication station, SAR host, FPGA, edge computing platform and micro POS system through lines to achieve high and low voltage conversion;

[0116] The edge computing platform is connected to the switch via a gigabit network port, is used to process optical payload images and SAR images in real time, and supports the deployment of the YOLOv11-TensorRT acceleration model; the edge computing platform is connected to the digital communication radio via a line to achieve remote data transmission;

[0117] The micro POS system is set close to the SAR antenna. The micro POS system transmits the position and attitude information of the UAV to the edge computing platform in real time through the internal data bus, so as to realize the coordinated work of SAR and optical payload 11;

[0118] The power module consists of a battery and a voltage conversion module. When the UAV does not provide power, it ensures the APOS can operate in a static state. When the UAV provides power, it stabilizes and rectifies the input power to provide a unified secondary power supply for each module.

[0119] The RF module consists of a transmitting channel, a receiving channel, and a frequency source. The transmitting channel performs up-conversion, filtering, and amplification on the digital intermediate frequency signal, enters the power amplifier module for power amplification, and then sends it to the transmitting antenna for external radiation. The receiving channel is responsible for amplifying, filtering, and gain controlling the echo signal received by the SAR antenna, and mixing it with the transmitting signal to obtain a difference frequency signal. The frequency source provides a coherent frequency reference source for each SAR extension.

[0120] The optical payload includes a dual optical pod for thermal imaging picture-in-picture switching, photo and video recording, target tracking and laser ranging.

[0121] Furthermore, the digital communication radio station includes a control system, a digital-to-analog converter, an analog-to-digital converter, and a timing and time service module; the control system is used to complete the control of the functions of each module, including receiving channel gain control, digital-to-analog converter and analog-to-digital converter working mode, and real-time processing flow control; the analog-to-digital converter module completes the acquisition of the receiving channel video signal; the analog-to-digital converter module generates a digital broadband intermediate frequency signal; the timing module provides a timing reference for the digital-to-analog converter, analog-to-digital converter and timing module; the timing module performs timing on the pulse repetition frequency and performs real-time image processing.

[0122] The beneficial effects brought about by the technical solution provided by the present invention include at least:

[0123] The present invention is based on the heterogeneous computing architecture of FPGA and edge computing platform, and is equipped with SAR host, POS, SAR antenna, optoelectronic payload and other hardware equipment. It cooperates with ground base stations to realize SAR real-time imaging processing, target detection, anomaly filtering, acquisition of positioning information, and realization of optoelectronic payload guidance and control, and parallel processing of collaborative detection and identification. It utilizes SAR's all-weather, all-day, wide-area detection capability that is not restricted by complex meteorological conditions and surface environment to provide optical payloads with efficient target pre-search and area guidance, breaking through the limitations of optical payloads' narrow field of view and dependence on good lighting conditions, and realizing unmanned system integration of wide-area scanning and close-range identification.

[0124] The present invention performs targeted demodulation processing on the echo signal, accurately optimizes the signal model, and effectively removes redundant information, laying a high-precision data foundation for subsequent core signal processing processes such as range compression and azimuth processing, improving the accuracy of target position solution and thereby enhancing target recognition accuracy.

[0125] Through echo signal processing and heterogeneous computing architecture design, the present invention effectively avoids the problems of bandwidth competition between massive SAR data and flight control instructions, and resource overload of a single hardware platform in traditional solutions, realizes the reasonable division and efficient execution of data processing tasks, and ensures the stability and real-time processing capability of the system during high-load operation. BRIEF DESCRIPTION OF THE DRAWINGS

[0126] Figure 1 A flow chart of the unmanned collaborative detection method based on SAR-guided optical payload provided by the present invention;

[0127] Figure 2 This is a schematic diagram of the overall structure of the unmanned collaborative detection system based on SAR-guided optical payload provided by the present invention;

[0128] Figure 3 A schematic diagram of the structure of the switch in the present invention;

[0129] Figure 4Schematic diagram of the structure of the optical payload in the present invention;

[0130] Figure 5 A schematic diagram of the three-dimensional structure of the unmanned collaborative detection system based on SAR-guided optical payload provided by the present invention;

[0131] Figure 6 This is a schematic diagram of the UAV flight operation process;

[0132] Figure 7 Flight routes for drones;

[0133] Figure 8 This is the imaging result after range compression and azimuth compression processing;

[0134] Figure 9 The SAR image detection results are displayed after the vehicle is deployed;

[0135] Figure 10 Display of pixel-geographic coordinate conversion results;

[0136] Figure 11 To guide and approach the effect, Figure 11 (a) is a schematic diagram of the actual position of the target; Figure 11 (b) is a schematic diagram of the first result of guiding the optical payload toward the actual position of the target; Figure 11 (c) is a schematic diagram of the second result of guiding the optical payload toward the actual target position.

[0137] Reference numerals:

[0138] 1. SAR antenna; 2. Digital communication radio; 3. Voltage converter; 4. SAR host; 5. Edge computing platform; 6. Micro POS system; 7. Digital communication radio antenna; 8. Switch; 9. Power module; 10. RF module; 11. Optical payload. DETAILED DESCRIPTION

[0139] To make the objectives, technical solutions and advantages of the present invention more clear, the embodiments of the present invention will be described in further detail below with reference to the accompanying drawings.

[0140] Example 1:

[0141] This embodiment will describe the unmanned collaborative detection method based on SAR guided optical payload of the present invention in more detail with reference to the accompanying drawings. Figure 1 As shown, the following steps are included:

[0142] S1: Scan the target area in a strip-like manner through SAR antenna 1 to obtain echo signals, and demodulate the echo signals to obtain demodulated echo signals;

[0143] S11: The radar signal is transmitted to the target area through SAR antenna 1. The echo signal reflected by the target and received is expressed as follows:

[0144] ;

[0145] in, It represents the echo signal received by SAR after being reflected by the target; is the target reflection coefficient, is a complex constant, t, Represent the slow time dimension and the fast time dimension respectively, is the moment when the center of the beam passes the target, and c is the propagation speed of the electromagnetic wave; Indicates the azimuth signal reception strength, where a represents the azimuth, which is the direction related to the SAR movement direction; Indicates the signal distance pulse envelope strength; is the slant range of SAR to the target at time t; exp() is the exponential function; is the center frequency, is the range modulation frequency; and They represent the signal range pulse envelope function and the azimuth signal receiving strength function respectively, and their expressions are as follows:

[0146] ;

[0147] in, is the pulse duration, P a is the azimuth antenna pattern function, rect() is the rectangular function, Indicates when When; sinc is the Sinker function; is the angle between the target and the beam centerline at time t; is the azimuth beamwidth, is the signal wavelength, is the antenna width in azimuth;

[0148] S12: Demodulate the echo signal to obtain a demodulated echo signal:

[0149] ;

[0150] in, represents the demodulated echo signal; j is the imaginary unit.

[0151] It should be noted that the Ku band used in the present invention has an operating frequency of 12.5-18 GHz. This band has low ground interference, high frequency, and is not easily affected by microwave radiation. The SAR system continuously transmits radar signals to the target area. The echo signals reflected by the target and received include carrier frequency information in addition to target information. Therefore, it is necessary to remove the carrier frequency component in the echo signal through demodulation. The demodulated SAR signal also becomes the SAR raw data. The present invention performs targeted demodulation processing on the echo signal, accurately optimizes the signal model, effectively strips off redundant information, and lays a high-precision data foundation for subsequent core signal processing processes such as range compression and azimuth processing, thereby improving the accuracy of target position solution and thus enhancing target recognition accuracy.

[0152] S2: Input the demodulated echo signal into FPGA, perform range and azimuth compression to realize imaging, and obtain SAR image;

[0153] The demodulated echo signal is input into the FPGA as the SAR raw data, and the range and azimuth compression are performed to realize imaging. The specific steps are as follows:

[0154] S21: distance compression;

[0155] Perform FFT on the distance direction of the demodulated echo signal and obtain the following expression:

[0156] ;

[0157] in, is the distance of the demodulated echo signal and the echo signal after FFT processing; G is the total gain including the scattering coefficient, which is set to 1; is the envelope of the range spectrum, is the distance frequency; is the slant range between the SAR and the target at the slow time η;

[0158] Range-matched filter The expression is:

[0159] ;

[0160] The echo signal after FFT processing is output through the range filter and then through IFFT to obtain the range compressed expression:

[0161] ;

[0162] in, is the echo signal after range compression; is the inverse fast Fourier transform; for IFFT expression of ;

[0163] S22: azimuth FFT;

[0164] SAR slant range to target :

[0165] ,

[0166] in, is the initial slant range, v is the SAR platform speed;

[0167] Since the beam is pointing close to the zero Doppler direction, and ,Will Approximately a parabola, the slant range of the SAR from the target , the calculation formula can be replaced by:

[0168] ;

[0169] When performing azimuth FFT, the time-frequency relationship is: ,in, is the azimuth frequency; is the frequency modulation in azimuth direction; is the movement speed of the SAR platform; is the wavelength of the SAR transmitted signal; the azimuth FFT expression is as follows:

[0170] ;

[0171] in, is the echo signal after FFT processing in azimuth; To perform fast Fourier transform on the time variable t; is the echo signal after range compression; is the azimuth frequency variable; for Frequency domain form of This is the distance migration compensation term, and its expression is as follows:

[0172] ;

[0173] S23: Perform distance migration correction;

[0174] The amount of distance migration that needs to be corrected , the expression is as follows:

[0175] ;

[0176] Range migration correction factor , the expression is as follows:

[0177] ;

[0178] The echo signal after range migration correction is , the expression is as follows:

[0179] ;

[0180] S24: azimuthal compression;

[0181] Azimuth matched filter The design is as follows:

[0182] ;

[0183] The echo signal after the azimuth matched filter output for:

[0184] ;

[0185] Finally, the final compressed signal is obtained through azimuth IFFT for:

[0186] ;

[0187] in, is the amplitude of the azimuthal impulse response;

[0188] After range compression and azimuth compression imaging processing, the final target has been corrected to At this point, target imaging processing is completed and a SAR image is obtained.

[0189] It should be noted that the echo signals received by the SAR antenna are sampled and then fed into the FPGA as input data. These echo signals are typically wide pulses containing reflection information from targets at varying distances. Therefore, range compression and azimuth focusing are required to achieve imaging. The present invention utilizes range migration correction and azimuth pulse compression techniques to achieve range compression and azimuth focusing, enabling real-time imaging.

[0190] S3: Transmit the SAR image to the edge computing platform via the COFDM data transmission link of data communication station 2;

[0191] It should be noted that the present invention effectively avoids the problems of bandwidth competition between massive SAR data and flight control instructions, resource overload of a single hardware platform, etc. in traditional solutions through echo signal processing and heterogeneous computing architecture (i.e., the FPGA and edge computing platform of the present invention), realizes the reasonable splitting and efficient execution of data processing tasks, and ensures the stability and real-time processing capability of the system during high-load operation.

[0192] S4: Perform INT8 quantization on the lightweight YOLOv11 model to obtain YOLO-TensorRT; perform target detection on the SAR image based on YOLO-TensorRT; determine whether a target is detected: if a target exists, proceed to S5; if no target is detected, return to S1;

[0193] It's important to note that in our object detection system, we used the lightweight YOLOv11 model and performed INT8 quantization on it. YOLOv11 was chosen because it offers high detection accuracy and fast inference speed for object detection tasks, meeting real-time requirements. INT8 quantization further reduces the model's computational workload and storage requirements, improving inference efficiency.

[0194] During the inference process, the YOLOv11 model extracts features and detects objects from the input image through a series of convolution, pooling, activation, and other operations. Specifically, the model outputs multiple prediction boxes, each of which contains the bounding box information of the target (center point coordinates x, y, width w, height h) and the confidence conf and target category probability p of the prediction box. class .

[0195] S5: Confidence screening of target detection results and geometric verification of the circumscribed circle to determine the final target;

[0196] After target detection based on YOLO-TensorRT, the model outputs a sequence of predicted boxes , where B is the prediction box sequence, is the i-th prediction box, and m is the total number of prediction boxes; ,in and is the coordinate of the center point of the i-th prediction box, and are the width and height of the i-th prediction box, is the confidence of the i-th prediction box, is the target category probability;

[0197] Through confidence screening and geometric verification of the circumscribed circle, abnormal targets are eliminated and the final target is determined. Specifically:

[0198] S51: confidence screening;

[0199] Set a confidence threshold , which is used to filter out prediction frames with higher confidence and obtain a valid prediction frame set; the filtering conditions are as follows:

[0200] ;

[0201] in, It is a set of valid prediction boxes obtained after confidence screening. Through this screening process, prediction boxes with low confidence and possible false positives can be eliminated.

[0202] S52: Circumscribed circle calculation and verification;

[0203] S521: Calculation of circumscribed circle;

[0204] For the valid prediction box set Each valid prediction box in , calculate the coordinates of the center of its circumscribed circle and radius ; Since the effective prediction box is a rectangle, the center of its circumscribed circle is the center point of the rectangle, and the radius is the distance from the center point to the vertex of the rectangle, specifically:

[0205] The formula for calculating the coordinates of the circle center is:

[0206] ;

[0207] Radius calculation formula:

[0208] Assume that a vertex of the valid prediction box is , according to the distance formula between two points, the radius for:

[0209] ;

[0210] S522: Geometric verification of circumscribed circle;

[0211] Detect the same target in N consecutive SAR images. Let the coordinates of the center of the circumscribed circle of the nth frame be , the radius is ; Calculate the mean of the center coordinates of N consecutive frames and the mean of the radius :

[0212] ;

[0213] Setting distance threshold and radius change threshold ; For each frame's center coordinates and radius , and calculate its deviation from the mean:

[0214] Center coordinate deviation :

[0215] ;

[0216] Radius deviation :

[0217] ;

[0218] like or , then the detection result of the frame is considered to be abnormal, and the valid prediction frame corresponding to the SAR image of the frame is removed;

[0219] S53: Final goal determined;

[0220] After confidence screening and circumscribed circle geometry verification, the target corresponding to the retained valid prediction box is the final target.

[0221] It should be noted that through confidence screening and geometric verification related to the circumscribed circle, abnormal targets can be effectively eliminated, the accuracy and reliability of target detection results can be improved, and more stable and accurate target information can be provided for subsequent processing steps.

[0222] S6: Combined with the SAR image, the coordinates of the center of the circumscribed circle of the final target are used as the input for pixel coordinate to geographic coordinate conversion, and the pixel coordinate to geographic coordinate conversion is performed, and the elevation correction is performed using the digital elevation model data to obtain the WGS-84 geographic coordinates of the final target;

[0223] S61: Geocoding Model

[0224] The system's GNSS / IMU data, combined with differential GNSS data from ground base stations, provides longitude and latitude information. The ground base stations continuously monitor satellite signals, calculate error corrections for satellite signals in the area, and broadcast them.

[0225] Assume that the longitude, latitude and altitude measured by GNSS on the UAV platform are , the error correction values ​​of longitude, latitude and altitude calculated by the ground base station are , then the position of the UAV platform after differential correction is:

[0226] ;

[0227] in, , , are the longitude, latitude and altitude of the platform after differential correction;

[0228] Combined with the SAR image, the coordinates of the center of the circumscribed circle of the final target are As input to the pixel coordinate to geographic coordinate conversion, establish pixel coordinates Mapping relationship to geographic space;

[0229] Assuming the platform position is the origin, calculate the initial latitude and longitude of the target using the spherical trigonometry formula :

[0230] ;

[0231] in, and are the initial longitude increment and latitude increment respectively; R e is the average radius of the earth; R is the slant range of the SAR to the target; φ is the azimuth of the target relative to the platform, according to the pixel coordinates Determination of the position in the SAR image and the geometric relationship of SAR imaging; is the cosine value of the UAV platform latitude after differential correction; is the beam pitch angle in SAR imaging; sin and cos are sine and cosine functions respectively;

[0232] The initial latitude and longitude coordinates are:

[0233] ;

[0234] S62: elevation correction;

[0235] The SRTM 30m resolution digital elevation model data is integrated to correct the terrain relief of the preliminary latitude and longitude coordinates. Assume that the target area elevation value obtained from the SRTM data is , the coordinate deviations caused by the terrain are and .

[0236] According to the geometric relationship, the coordinate deviation can be approximately calculated:

[0237] ;

[0238] Finally, we get the precise WGS-84 geographic coordinates (lon, lat):

[0239] .

[0240] S7: The edge computing platform 5 transmits the WGS-84 geographic coordinates of the final target to the ground station via the ground base station. The drone receives the approach flight path command forwarded by the ground station via the ground base station, adjusts its heading and altitude, and flies toward the target area. The optical payload activates high-definition imaging mode to verify the target details.

[0241] S71: Boot instruction generation;

[0242] The edge computing platform 5 transmits the converted target coordinates (e.g., in JSON format {"lon":116.4034,"lat":39.9234,"radius":5.0}) to the ground station via the ground base station. The ground base station plays a key role in the data transmission process. It uses an interference-resistant digital link (such as COFDM) to ensure stable data transmission in complex electromagnetic environments.

[0243] After the ground station receives the WGS-84 geographic coordinates of the final target, it combines the preset safety radius (r is the target feature radius), through the path planning algorithm (such as Algorithm) generates the approach flight path of the UAV. In the algorithm, the cost function f(e)=g(e)+h(e) is calculated for each node, where g(e) is the actual cost from the starting node to the current node e, and h(e) is the heuristic estimated cost from the current node e to the target node;

[0244] Let the starting node be S, the target node be M, the current node be e, and the distance between any two nodes be It can be calculated based on the latitude and longitude coordinates:

[0245] ;

[0246] Among them, (lon1, lat1) is the longitude and latitude coordinates of node e1; (lon2, lat2) is the longitude and latitude coordinates of node e2, e1 and e2 are used to represent any two nodes;

[0247] Then we can get that g(e) is the sum of the distances of all edges from the starting node S to the node e; h(e) uses the straight-line distance as a heuristic estimate:

[0248] h(e)=d(e,M);

[0249] S72: Task closure;

[0250] The drone receives commands forwarded by the ground station via the ground base station, adjusts its heading and altitude, and flies toward the target area. The optical payload activates high-definition imaging mode (such as 30x zoom) to verify target details.

[0251] The ground base station also plays a role in dynamic compensation during the whole process. Since the complex electromagnetic environment may cause attenuation and interference of communication signals, the ground base station monitors the quality of the communication link in real time, such as signal strength. , bit error rate BER and other indicators.

[0252] Let the signal strength threshold be , the bit error rate threshold is .when or When the ground base station dynamically switches channels or enhances signal power, for example, by adjusting the gain of the transmitting antenna To boost signal power:

[0253] ;

[0254] in is the original transmit power, which is dynamically adjusted To ensure the stability of the communication link, thereby ensuring the stability of command transmission and pod return image, forming a complete mission closed loop of "discovery-guidance-verification".

[0255] Example 2:

[0256] This embodiment will describe the unmanned collaborative detection system based on SAR guided optical payload of the present invention in more detail with reference to the accompanying drawings. Figure 1 As shown, the system includes: a drone, a SAR, a digital communication radio 2, a voltage converter 3, an edge computing platform 5, a micro POS system 6, a power module 9, a radio frequency module 10 and an optical payload 11;

[0257] The SAR is installed on a drone and includes a SAR antenna 1 and a SAR host 4. The SAR antenna 1 serves as the SAR's front-end component, transmitting and receiving radar signals. The SAR host 4 utilizes an FPGA motherboard, provides Gigabit Ethernet connectivity to an edge computing platform 5, and integrates a radar waveform generation module that supports dynamic pulse repetition frequency adjustment. The SAR generates SAR images after processing the signals received by the SAR antenna 1 through the SAR host's FPGA. Furthermore, the SAR antenna of the present invention features high gain and directional beam characteristics, operates in the X-band (8-12 GHz) frequency band, has a gain of ≥25 dBi, and is horizontally polarized. It is mounted on the exterior of the drone's pod and utilizes lightweight composite materials to reduce wind resistance. The SAR host, the core signal processing unit, utilizes a Xilinx Kintex-7XC7K325T FPGA motherboard, capable of pulse compression, MTI filtering, and data scheduling.

[0258] The digital communication radio 2 is equipped with a digital communication radio antenna 7 for data transmission between the SAR and the ground station. Furthermore, the digital communication radio is located on the right side of the integrated platform, facilitating wireless signal transmission and reception while minimizing internal module interference. The digital communication radio 2 supports the L-band, a maximum transmission range of 50 km, QPSK modulation, a built-in encryption module, and RS422 / Ethernet interfaces.

[0259] The voltage converter 3 is located below the digital communication station 2 and is connected to the digital communication station 2, SAR host 4, FPGA, edge computing platform, and micro POS system 6 via lines to achieve high-to-low voltage conversion. Furthermore, the voltage converter is located near the power module to facilitate input power processing and voltage stabilization, ensuring stable operation of each module.

[0260] The edge computing platform 5 is connected to the switch 8 via a gigabit Ethernet port, enabling real-time processing of optical payload images and SAR images, and supporting the deployment of the YOLOv11-TensorRT acceleration model. The edge computing platform 5 is also connected to the data communication radio 2 via a line to enable remote data transmission. It should be noted that the edge computing platform (NVIDIA Jetson Orin NX) in this invention supports channel pruning and INT8 quantization.

[0261] The micro-POS system 6 is located near the SAR antenna 1. It transmits the drone's position and attitude information to the edge computing platform 5 in real time via an internal data bus, enabling the coordinated operation of the SAR and optical payload 11. It should be noted that the micro-POS system (positioning and orientation system) integrates modules such as GNSS, IMU, and inertial navigation. By fusing data from these components, it can more accurately obtain position and attitude information related to radar signals. The GNSS (Global Navigation Satellite System) primarily provides positioning information, providing the foundation for the system to determine the target's geographic location. The IMU (Inertial Measurement Unit) measures information such as attitude angles. In this embodiment, the micro-POS system's parameters are set as follows: GNSS positioning accuracy of ±0.1m, attitude angle accuracy of ±0.1, and an output frequency of 100Hz. The integrated IMU and inertial navigation are synchronized with the radar's PRF via a PPS signal. Furthermore, the micro-POS system acquires position and attitude information related to radar signals, providing a precise spatiotemporal reference for subsequent processing.

[0262] The power module 9 is composed of a battery and a voltage conversion module. When the UAV does not provide power, it ensures the power supply requirements of the APOS (Airborne Position and Orientation System) onboard position and attitude system in a stationary state. When the UAV provides power, it stabilizes and rectifies the input power to provide a unified secondary power supply for each module. After the voltage converter 3 stabilizes and rectifies the input power, it is connected to the edge computing platform 5, the digital communication radio 2, the micro POS system 6 and other modules through lines to ensure the stable operation of each module.

[0263] The RF module 10 consists of a transmitting channel, a receiving channel, and a frequency source. The transmitting channel performs up-conversion, filtering, and amplification on the digital intermediate frequency signal, enters the power amplifier module for power amplification, and then sends it to the transmitting antenna for external radiation. The receiving channel is responsible for amplifying, filtering, and gain controlling the echo signal from the receiving antenna, and mixing it with the transmitting signal to obtain a difference frequency signal. The frequency source provides a coherent frequency reference source for each SAR extension.

[0264] The optical payload 11 includes a dual-optical pod, which is used for thermal image picture-in-picture switching, photo and video recording, target tracking, and laser ranging. Specifically, in this embodiment, the pod uses the Q30TIRMplus, a high-precision, three-axis stabilized dual-optical pod equipped with a 2.13-megapixel Sony camera with 30x optical zoom, a 25mm lens, a 640x480 resolution thermal imager, and a 2000-meter rangefinder. The dual-optical pod supports visible light zoom, thermal image picture-in-picture switching, multi-color palette switching, photo and video recording, target tracking, thermal image electronic zoom, and laser ranging.

[0265] Furthermore, the digital communication radio 2 includes a control system, an analog to digital converter (ADC), an analog to digital converter (DAC), a timing and time service module, and a recording module; the control system is used to control the functions of each module, including receiving channel gain control, ADC and DAC working modes, and real-time processing flow control; the ADC module completes the acquisition of the receiving channel video signal, and the DAC module generates a digital broadband intermediate frequency signal; the timing module provides a timing reference for the ADC, DAC and timing module, and the timing module performs time service on the pulse repetition frequency and performs real-time image processing.

[0266] It should be noted that the present invention is based on the heterogeneous computing architecture of FPGA and edge computing platform, and is equipped with SAR host, POS, SAR antenna, optoelectronic payload and other hardware equipment, and cooperates with ground base stations to realize SAR real-time imaging processing, target detection, anomaly filtering, acquisition of positioning information, and realization of optoelectronic payload guidance and control, collaborative detection and identification parallel processing. It utilizes SAR's all-weather, all-day, wide-area detection capability that is not restricted by complex meteorological conditions and surface environment to provide optical payloads with efficient target pre-search and area guidance, breaking through the limitations of optical payloads with narrow field of view and dependence on good lighting conditions, and realizing unmanned system integration of wide-area scanning and close-range identification.

[0267] Simulation test

[0268] The effectiveness of the proposed method is evaluated in a simulation scenario. Figure 6A rotorcraft drone was used as a test platform. The ground-based remote control device and Computer 2 formed the drone's takeoff and flight path controller, while the base station recorded the drone's flight path. Once the drone took off and stabilized on its flight path, Computer 1 issued commands via the ground and airborne radio data links to control the SAR host's power on and off, collect raw SAR data, and perform subsequent data processing.

[0269] Schematic diagram of the UAV flight operation process Figure 6 As shown in the figure, the SAR system is first activated, scanning the selected area in a strip-like manner. This method allows for a large-scale, rapid search of the designated area, initially determining the approximate location of a suspicious (or interesting) target. This location information is then transmitted to the pod, guiding it in the designated direction for further, detailed observation.

[0270] SAR antenna strip scanning

[0271] The drone flies according to the predetermined route. The flight route is as follows: Figure 7 As shown in the figure, the drone ascends vertically from point A and enters the flight path after reaching an altitude of 250 meters. The SAR system is turned on and operates in strip scanning mode, viewing the right side. Before each corner, the SAR system is turned off and not on until the drone stabilizes on the next straight line. The SAR system is then turned on again. The mission is complete when the drone reaches point B.

[0272] FPGA real-time imaging processing

[0273] The collected raw SAR data is input into FPGA, processed by range pulse compression and azimuth pulse compression, and imaged to obtain the imaging results as shown below: Figure 8 shown.

[0274] Data link transmission

[0275] The SAR image obtained by FPGA imaging processing is transmitted in real time to the edge terminal platform (Jetson Orin) through the COFDM data transmission link in the form of SAR image data blocks (JPEG format).

[0276] Real-time object detection based on YOLO-TensorRT

[0277] The SAR image input to the edge terminal is processed by the YOLO-TensorRT algorithm to obtain the result after algorithm processing.

[0278] Confidence screening and geometric verification

[0279] like Figure 9As shown in the figure, after target detection, a series of target prediction frames can be obtained. After these prediction frames are processed through confidence screening and circumscribed circle geometry verification, the final target prediction frame can be screened out.

[0280] Pixel to Geographic Coordinate Conversion

[0281] like Figure 10 As shown in the figure, after confidence screening and circumscribed circle stability verification, the objects corresponding to the retained prediction boxes are the final valid objects. The coordinates of the circumscribed circles of these objects will be used as input for the subsequent pixel coordinate to geographic coordinate conversion, thereby obtaining the actual geographic coordinate position of the object.

[0282] Optical payload guidance and approach flight

[0283] like Figure 11 As shown in Figure 11 As shown in (a), guide the pod towards the designated position as Figure 11 (b) and Figure 11 (c) and obtain more detailed information of the target, such as Figure 11 shown.

[0284] It should be noted that the serial numbers of the above-mentioned embodiments of the present invention are for descriptive purposes only and do not represent the advantages or disadvantages of the embodiments. In addition, the terms "including", "comprising" or any other variations thereof are intended to cover non-exclusive inclusion, so that a process, device, article or method comprising a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, device, article or method. In the absence of further restrictions, an element defined by the sentence "including a ..." does not exclude the presence of other identical elements in the process, device, article or method comprising the element.

Claims

1. An unmanned collaborative detection method based on SAR-guided optical payload, characterized in that: include: S1: The SAR antenna is used to scan the target area in strips to obtain echo signals, and the echo signals are demodulated to obtain demodulated echo signals; S2: Input the demodulated echo signal into FPGA, perform range and azimuth compression to realize imaging, and obtain SAR image; S3: Transmit the SAR image to the edge computing platform via the COFDM data transmission link of the data communication radio; S4: Perform INT8 quantization on the lightweight YOLOv11 model to obtain YOLO-TensorRT; Perform target detection on the SAR image based on YOLO-TensorRT; determine whether the target is detected: if there is a target, proceed to S5; if no target is detected, return to S1; S5: Confidence screening of target detection results and geometric verification of the circumscribed circle to determine the final target; S6: Combined with the SAR image, the coordinates of the center of the circumscribed circle of the final target are used as the input for pixel coordinate to geographic coordinate conversion, and the pixel coordinate to geographic coordinate conversion is performed, and the elevation correction is performed using the digital elevation model data to obtain the WGS-84 geographic coordinates of the final target; S7: The edge computing platform transmits the WGS-84 geographic coordinates of the final target to the ground station through the ground base station. The UAV receives the approach flight path instructions forwarded by the ground station through the ground base station, adjusts the heading and altitude, and flies towards the target area; the optical payload starts the high-definition imaging mode to verify the target details.

2. The unmanned collaborative detection method based on SAR guided optical payload according to claim 1, characterized in that: Said S1 comprises: S11: The radar signal is transmitted to the target area through the SAR antenna. The echo signal reflected by the target and received is expressed as follows: ； in, It represents the echo signal received by SAR after being reflected by the target; is the target reflection coefficient, is a complex constant, t, Represent the slow time dimension and the fast time dimension respectively, is the moment when the center of the beam passes the target, and c is the propagation speed of the electromagnetic wave; Indicates the azimuth signal reception strength, where a represents the azimuth, which is the direction related to the SAR movement direction; Indicates the signal distance pulse envelope strength; is the slant range of SAR to the target at time t; exp() is the exponential function; is the center frequency, is the range modulation frequency; and They represent the signal range pulse envelope function and the azimuth signal receiving strength function respectively, and their expressions are as follows: ； in, is the pulse duration, P a is the azimuth antenna pattern function, rect() is the rectangular function, Indicates when When; sinc is the Sinker function; is the angle between the target and the beam centerline at time t; is the azimuth beamwidth, is the signal wavelength, is the antenna width in azimuth; S12: Demodulate the echo signal to obtain a demodulated echo signal: ； in, represents the demodulated echo signal; j is the imaginary unit.

3. The unmanned collaborative detection method based on SAR guided optical payload according to claim 2, characterized in that: The S2 includes: The demodulated echo signal is input into the FPGA as the SAR raw data, and the range and azimuth compression are performed to realize imaging. The specific steps are as follows: S21: distance compression; Perform FFT on the distance direction of the demodulated echo signal and obtain the following expression: ； in, is the distance of the demodulated echo signal and the echo signal after FFT processing; G is the total gain including the scattering coefficient, which is set to 1; is the envelope of the range spectrum, is the distance frequency; is the slant range between the SAR and the target at the slow time η; Range-matched filter The expression is: ； After FFT processing, the echo signal is output through the range filter and then through IFFT to obtain the range compressed expression. : ； in, is the inverse fast Fourier transform; for IFFT expression of ; S22: azimuth FFT; SAR slant range to target : ； in, is the initial slant range, v is the SAR platform speed; Since the beam is pointing close to the zero Doppler direction, and ,Will Approximately a parabola, the slant range of the SAR from the target , the calculation formula can be replaced by: ； When performing azimuth FFT, the time-frequency relationship is: ,in, is the azimuth frequency; is the frequency modulation in azimuth direction; is the movement speed of the SAR platform; is the wavelength of the SAR transmitted signal; the azimuth FFT expression is as follows: ； in, is the echo signal after FFT processing in azimuth; To perform fast Fourier transform on the time variable t; is the echo signal after range compression; is the azimuth frequency variable; for Frequency domain form of is the distance migration compensation term, and its expression is as follows: ； S23: Perform distance migration correction; The amount of distance migration that needs to be corrected , the expression is as follows: ； Range migration correction factor , the expression is as follows: ； The echo signal after range migration correction is , the expression is as follows: ； S24: Azimuthal compression.

4. The unmanned collaborative detection method based on SAR guided optical payload according to claim 3, characterized in that: The S24 azimuth pulse compression includes: Azimuth matched filter The design is as follows: ； The echo signal after the azimuth matched filter output for: ； Finally, the final compressed signal is obtained through azimuth IFFT for: ； in, is the amplitude of the azimuthal impulse response; After range compression and azimuth compression imaging processing, the final target has been corrected to At this point, target imaging processing is completed and a SAR image is obtained.

5. The unmanned collaborative detection method based on SAR guided optical payload according to claim 4, characterized in that: The S5 includes: After target detection based on YOLO-TensorRT, the model outputs a sequence of predicted boxes , where B is the prediction box sequence, is the i-th prediction box, and m is the total number of prediction boxes; ,in and is the coordinate of the center point of the i-th prediction box, and are the width and height of the i-th prediction box, is the confidence of the i-th prediction box, is the target category probability; Through confidence screening and geometric verification of the circumscribed circle, abnormal targets are eliminated and the final target is determined. Specifically: S51: confidence screening; Set a confidence threshold , which is used to filter out prediction frames with higher confidence and obtain a valid prediction frame set; the filtering conditions are as follows: ； in, It is the set of valid prediction boxes obtained after confidence screening; S52: Circumscribed circle calculation and verification; S521: Calculation of circumscribed circle; For the valid prediction box set Each valid prediction box in , calculate the coordinates of the center of its circumscribed circle and radius ; Since the effective prediction box is a rectangle, the center of its circumscribed circle is the center point of the rectangle, and the radius is the distance from the center point to the vertex of the rectangle, specifically: The formula for calculating the coordinates of the circle center is: ； Radius calculation formula: Assume that a vertex of the valid prediction box is , according to the distance formula between two points, the radius for: ； S522: Geometric verification of circumscribed circle; Detect the same target in N consecutive SAR images. Let the coordinates of the center of the circumscribed circle of the nth frame be , the radius is ; Calculate the mean of the center coordinates of N consecutive frames and the mean of the radius : ； Setting distance threshold and radius change threshold ; For each frame's center coordinates and radius , and calculate its deviation from the mean: Center coordinate deviation : ； Radius deviation : ； like or , then the detection result of the frame is considered to be abnormal, and the valid prediction frame corresponding to the SAR image of the frame is removed; S53: Final goal determined; After confidence screening and circumscribed circle geometry verification, the target corresponding to the retained valid prediction box is the final target.

6. The unmanned collaborative detection method based on SAR guided optical payload according to claim 5, characterized in that: The S6 includes: S61: Geocoding model; Assume that the longitude, latitude and altitude measured by GNSS of the UAV platform are , the error correction values ​​of longitude, latitude and altitude calculated by the ground base station are , then the position of the UAV platform after differential correction is: ； in, , , are the longitude, latitude and altitude of the platform after differential correction; Combined with the SAR image, the coordinates of the center of the circumscribed circle of the final target are As input to the pixel coordinate to geographic coordinate conversion, establish pixel coordinates Mapping relationship to geographic space; Assuming the platform position is the origin, calculate the initial latitude and longitude of the target using the spherical trigonometry formula : ; in, and are the initial longitude increment and latitude increment respectively; R e is the average radius of the earth; R is the slant range of the SAR to the target; φ is the azimuth of the target relative to the platform, according to the pixel coordinates Determination of the position in the SAR image and the geometric relationship of SAR imaging; is the cosine value of the UAV platform latitude after differential correction; is the beam pitch angle in SAR imaging; sin and cos are sine and cosine functions respectively; The initial latitude and longitude coordinates are: ； S62: elevation correction; The SRTM 30m resolution digital elevation model data was integrated to correct the terrain relief of the preliminary latitude and longitude coordinates. Specifically: Assume that the target area elevation value obtained from SRTM data is , the coordinate deviations caused by the terrain are and ; According to the geometric relationship, the coordinate deviation can be approximately calculated: ； Among them, tan is the tangent function; Get the WGS-84 geographic coordinates (lon, lat) of the final target: 。 7. The unmanned collaborative detection method based on SAR guided optical payload according to claim 6, characterized in that: The S7 includes: S71: The edge computing platform transmits the WGS-84 geographic coordinates of the final target to the ground station via the ground base station. After receiving the WGS-84 geographic coordinates of the final target, the ground station generates the approach flight path of the UAV using a path planning algorithm. S72: The UAV receives the approach flight path command forwarded by the ground station via the ground base station, adjusts its heading and altitude, and flies toward the target area. The optical payload activates high-definition imaging mode to verify target details. Ground base stations monitor the signal strength of communication links in real time and bit error rate BER; Let the signal strength threshold be , the bit error rate threshold is ; when or When the signal is transmitted, the ground base station dynamically switches channels or enhances signal power.

8. An unmanned collaborative detection system based on SAR-guided optical payload, to implement the unmanned collaborative detection method based on SAR-guided optical payload according to any one of claims 1 to 7, characterized in that: The system includes: a drone, a SAR, a data communication radio, a voltage converter, an edge computing platform, a micro POS system, a power module, a radio frequency module, and an optical payload; The SAR is installed on a drone and includes a SAR antenna and a SAR host. The SAR antenna serves as the front-end component of the SAR and is used to transmit and receive radar signals. The SAR host is composed of an FPGA motherboard, provides a Gigabit Ethernet connection to the edge computing platform, integrates a radar waveform generation module, and supports dynamic adjustment of the pulse repetition frequency. The SAR obtains a SAR image after the signal received by the SAR antenna is processed by the FPGA of the SAR host. The digital communication station is provided with a digital communication station antenna for realizing data transmission between the SAR and the ground station; The voltage converter is located below the digital communication station, and is connected to the digital communication station, SAR host, FPGA, edge computing platform and micro POS system through lines to achieve high and low voltage conversion; The edge computing platform is connected to the switch via a gigabit network port, is used to process optical payload images and SAR images in real time, and supports the deployment of the YOLOv11-TensorRT acceleration model; the edge computing platform is connected to the digital communication radio via a line to achieve remote data transmission; The micro POS system is set up close to the SAR antenna. The micro POS system transmits the position and attitude information of the UAV to the edge computing platform in real time through the internal data bus, realizing the collaborative work of SAR and optical payload. The power module consists of a battery and a voltage conversion module. When the UAV does not provide power, it ensures the APOS can operate in a static state. When the UAV provides power, it stabilizes and rectifies the input power to provide a unified secondary power supply for each module. The RF module consists of a transmitting channel, a receiving channel, and a frequency source. The transmitting channel performs up-conversion, filtering, and amplification on the digital intermediate frequency signal, enters the power amplifier module for power amplification, and then sends it to the transmitting antenna for external radiation. The receiving channel is responsible for amplifying, filtering, and gain controlling the echo signal received by the SAR antenna, and mixing it with the transmitting signal to obtain a difference frequency signal. The frequency source provides a coherent frequency reference source for each SAR extension. The optical payload includes a dual optical pod for thermal imaging picture-in-picture switching, photo and video recording, target tracking and laser ranging.

9. The unmanned collaborative detection system based on SAR guided optical payload according to claim 8, characterized in that: The digital communication radio station includes a control system, a digital-to-analog converter, an analog-to-digital converter, and a timing and time service module; the control system is used to control the functions of each module, including receiving channel gain control, the working mode of the digital-to-analog converter and the analog-to-digital converter, and real-time processing flow control; the analog-to-digital converter module completes the acquisition of the receiving channel video signal; The analog-to-digital converter module generates a digital broadband intermediate frequency signal; the timing module provides a timing reference for the digital-to-analog converter, the analog-to-digital converter and the timing module; the timing module performs timing on the pulse repetition frequency and processes real-time images.

Images