Missile-target intersection parameter measurement and real-time resolving method and system based on area array laser radar

Abstract

The invention belongs to the field of missile-target rendezvous parameter measurement, and provides a missile-target rendezvous parameter measurement and real-time resolving system and method based on an area array laser radar. The observation end utilizes a mooring unmanned aerial vehicle to carry an area array laser radar, a visual sensor and a positioning and attitude determination unit to observe a flight area and a target and collect data, and then utilizes an industrial personal computer unit to carry out storage and data preprocessing on collected target point cloud data, image data and observation end position attitude data. A roughly spliced three-dimensional environment model is obtained, and the original data and the preprocessed point cloud data are transmitted back to the ground end display control unit and the missile target rendezvous parameter resolving unit in real time through the wired communication unit; and the missile-target rendezvous parameter resolving unit is used for performing data processing operations such as point cloud denoising, target extraction, point cloud registration and axis extraction on the point cloud data, resolving missile-target rendezvous parameters such as a movement track, rendezvous speed and position information of the flying target, and transmitting a resolving result to the display control unit.

Description

Technical Field

[0001] The present invention belongs to the field of missile-target intersection parameter measurement, relates to the field of flash lidar technology application, and specifically relates to a missile-target intersection parameter measurement and real-time solution method and system based on flash lidar. Technical Background

[0002] The measurement of missile-target intersection parameters mainly completes the measurement of parameters such as ballistic inclination, ballistic deviation, projectile posture, and relative position of projectile and target. The measurement results can provide important data support for the assessment of projectile terminal guidance accuracy, evaluation of ignition-warfare coordination effectiveness, and judgment of killing effects, which is of great significance.

[0003] At present, the methods for measuring the intersection parameters of projectiles and targets mainly include direct measurement and non-contact measurement based on optical sensors. The direct measurement method is to install sensors such as gyroscopes, lasers and GNSS inside the projectile. The accuracy of the sensors directly affects the measurement accuracy of the intersection parameters of the projectile and target. Therefore, this method has high requirements on the performance of the sensors, and the installation of the sensors will destroy the inherent structure of the projectile. The observation method based on optical sensors can only collect two-dimensional image information of the target with a single device, and cannot obtain the distance information of the target. It requires more than two devices to perform intersection measurement to determine the position and angle relationship between the projectile and the target. In addition, high-speed cameras or photoelectric theodolites are expensive, and their large size and weight also limit their platform selection and equipment layout. These two measurement methods have certain limitations in terms of real-time data solution, multi-platform adaptability, parameter measurement accuracy and equipment layout, and are difficult to meet the measurement needs of modern strong mobility and complex environments.

[0004] As a new type of high-precision real-time three-dimensional remote sensing measurement technology, the array laser radar has the advantages of high resolution, high frame rate, compact structure, light weight and portability, which makes it flexibly applicable to various complex environments and various measurement platforms. It has many natural advantages in the measurement of projectile-target intersection parameters, such as high measurement accuracy, strong anti-electromagnetic interference ability and convenient deployment. On the basis of fully analyzing the measurement requirements of projectile-target intersection parameters and the performance parameters of the array laser radar, the array laser radar is upgraded and reconstructed, and the array laser radar measurement system and real-time data solution method for projectile-target intersection parameter measurement are established to replace the traditional direct measurement method and optical observation method, so as to improve the measurement accuracy of projectile-target intersection parameters and the efficiency of real-time parameter solution, and provide data support for weapon performance assessment test. Summary of the invention

[0005] In view of the limitations of the existing methods for measuring projectile-target intersection parameters, the main purpose of the present invention is to provide a method and system for measuring and real-time solving projectile-target intersection parameters based on area array laser radar, which aims to replace the existing projectile-target intersection parameter measurement method, improve the measurement accuracy of projectile-target intersection parameters and reduce the measurement cost.

[0006] To achieve the above object, the technical solution of the present invention is as follows:

[0007] A missile-target intersection parameter measurement and real-time solution system based on array laser radar includes an observation end and a ground workstation end; the observation end includes an array laser radar unit, a visual sensor unit, a positioning and attitude system unit, an industrial computer unit, a wired communication unit and a power adapter module, and the ground workstation end includes a display control unit, a missile-target intersection parameter solution unit, a wired communication unit and a tethered power supply unit.

[0008] The array laser radar unit is connected to the industrial computer unit, the positioning and attitude system unit, the wired communication unit and the power adapter module, collects the true color three-dimensional point cloud data of the target movement in real time, and transmits the data to the industrial computer in real time for storage, forwarding and data preprocessing.

[0009] The visual sensor unit is connected to the industrial computer unit and the power adapter module, takes photos of the target area at a certain frame rate, and transmits the collected image data to the industrial computer for storage, forwarding and image data preprocessing.

[0010] The positioning and attitude determination system unit is composed of an IMU (inertial measurement unit) module and a Beidou positioning module. The IMU module is used to obtain the speed and attitude information of the observation end, and the Beidou positioning module is used to obtain the longitude, latitude and elevation information of the observation end. The collected information is transmitted to the industrial computer unit in real time for data storage and forwarding.

[0011] The industrial computer unit is composed of a data storage module and a data preprocessing module, and the array laser radar unit, the visual sensor unit and the positioning and attitude system unit are all connected to the industrial computer unit; the data storage module is used to store the original point cloud data of the target movement collected by the array laser radar unit, the target area image data collected by the visual sensor, and the observation system position and attitude data collected by the positioning and attitude system unit, and the data preprocessing module is used to jointly solve the original point cloud data, image data and position and attitude data to achieve rough splicing of the point cloud data and generate a large-scale three-dimensional environment model.

[0012] The wired communication unit includes a wired communication unit at the observation end and a wired communication unit at the ground end. The two wired communication units transmit data through a network cable and are responsible for the communication transmission services of the observation end and the ground end respectively. The wired communication unit at the observation end is connected to the industrial computer, and transmits the position and posture data collected by the positioning and attitude system unit, the image data collected by the visual sensor unit, and the point cloud data pre-processed by the industrial computer unit back to the ground end in real time; the wired communication unit at the ground end transmits the data to the display control unit and the missile-target intersection parameter calculation unit in real time, and transmits the observation command issued by the operator back to the observation end.

[0013] The power adapter module is connected to the tethered power supply unit at the ground end through a tethered cable, providing safe and stable power supply support for the area array laser radar unit, visual sensor unit, industrial computer unit and positioning and attitude determination system unit at the observation end.

[0014] The display control unit shares data with the missile-target intersection parameter solving unit, and displays two parts of data information transmitted from the observation end and the missile-target intersection parameter solving unit on the ground end. One part displays the pre-processed point cloud data and the observation system position, posture and other information sent back in real time by the observation end, and the other part displays the parameter settlement result sent back by the missile-target intersection parameter solving unit, so as to provide decision-making information for the operator to perform supplementary measurement and encrypted measurement.

[0015] The tethered power supply unit is connected to the power adapter module via a tethered cable to provide safe and stable power supply support for the observation end and the ground end.

[0016] The missile-target intersection parameter calculation unit is composed of a measured data import and simulation data generation module, a point cloud image enhancement and filtering module, a target intersection parameter calculation module and a measurement accuracy evaluation module, and is connected to a data wired communication unit and a display control unit. The measured data import and simulation data generation module imports the measured data collected by the array laser radar unit in real time, and uses the STL model of the typical scene model library imported from VTK, mainly including airports, ports, and land buildings; the point cloud image enhancement and filtering module uses a dual-modal information fusion algorithm and a RANSAC algorithm to extract point cloud targets and detect the ground, and uses a K-means clustering algorithm to perform point cloud denoising; the target intersection parameter calculation module estimates the target's motion trajectory through polynomial fitting based on the extracted target information and the axis equation, and calculates the intersection parameters such as motion posture, target terminal speed, and intersection posture angle; the measurement accuracy evaluation module uses the average Euclidean distance to define the error result between the calculated target information and the simulation information, and realizes the comparative analysis and accuracy evaluation functions based on the comparison between the calculated target information and the simulation information.

[0017] A method for measuring and real-time solving missile-target intersection parameters based on an array laser radar adopts the above system and includes the following steps:

[0018] Step A: All measurement units at the observation end are loaded onto the tethered drone, and the drone is flown to the planned location for hovering observation. The array laser radar and visual sensor collect the target 3D point cloud data and target area image data in real time at a fixed frame rate and viewing angle. The positioning and attitude system unit collects the position and attitude data of the observation end. The collected data are transmitted to the industrial computer unit for data storage, forwarding and preprocessing.

[0019] Step B: The data preprocessing module in the industrial computer unit jointly solves the collected observation end position and posture data, image data and original point cloud data, completes the rough stitching of the point cloud data, and solves the observation end position and posture changes to achieve self-positioning and obstacle avoidance navigation of the UAV.

[0020] Step C: The wired communication unit transmits the rough stitching point cloud data and the observation end position and posture data pre-processed by the industrial computer to the display control unit of the ground workstation in real time. The display control unit transmits the data to the missile-target intersection parameter solution unit for real-time solution, and sends the solution result to the display control unit. The operator controls the position of the tethered UAV and the direction of the adjustable self-stabilizing gimbal on the aircraft according to the real-time solution result, and performs additional measurement of key targets.

[0021] Step D: The missile-target intersection parameter calculation unit performs real-time calculation on the returned data, including point cloud denoising, target extraction, registration, and trajectory fitting. After the calculation is completed, the calculation result is sent back to the display control unit, as follows:

[0022] D1: In order to solve the problem that the background point cloud at the edge is easily extracted when extracting the target point cloud, the present invention adopts the K-means clustering algorithm to set the radius threshold, compare the Euclidean distance of the randomly selected points, and delete all points greater than the radius threshold to achieve the denoising function, so as to remove the noise points other than the target. For complex and irregular point cloud models, the k-means clustering algorithm has a good clustering effect, and its algorithm time complexity is relatively low.

[0023] D2: After the original point cloud data segmentation is completed, there are still a large number of point clouds scanned to the ground, so the ground point cloud needs to be identified and removed, and only the obstacle point cloud is retained. The present invention uses the RANSAC algorithm to fit the plane and divides the original data into two parts: internal points and external points. The internal points satisfy the plane model and belong to the ground points, while the external points do not satisfy this model and belong to obstacle points.

[0024] D3: Image data lacks depth information when describing target motion. Point cloud data has a large amount of data and a long calculation time when extracting small targets in complex environments. The present invention adopts a motion target extraction method with dual-modal information fusion, establishes a coordinate transformation model between a color camera and an array laser radar to achieve pixel-level matching of the two sensors, and fuses the image and point cloud at the same time; specifically, first, an improved background difference algorithm is used to extract the image motion target, a mean filter is used to remove image noise, and related algorithms in the OpenCV library are used to extract the image target contour, and finally the corresponding point cloud is selected according to the position of the image target; then, after the point cloud is projected to the image pixel coordinate system, the position of the target three-dimensional point cloud in the original point cloud can be obtained by observing the position of the target projection point cloud in the overall projection point cloud.

[0025] D4: After completing the target point cloud data segmentation, the present invention uses the ICP (iterative closest point) algorithm to realize the registration of the target point cloud and the point cloud model, obtain the posture conversion matrix, and estimate the posture information of the missile.

[0026] D5: Based on the rotated point cloud model, normal fitting is performed first, and then the cylindrical surface fitting of the point cloud target is performed using the least squares method to realize the extraction of the axis.

[0027] D6: Based on the extracted target information and axis equation, the polynomial fitting method is used to estimate the actual motion trajectory of the target. The trajectory prediction method based on Gaussian mixture model is used to model complex motion patterns, calculate the probability distribution of different motion patterns, divide the trajectory data into different components, and use Gaussian process regression to predict the most likely motion posture of the moving object.

[0028] D7: The time, position and angle information of the target can be obtained based on the point cloud data collected by the area array laser radar. The target angle change is abstracted as a polynomial of the target's rotation angle in the xOy and xOz planes with respect to time. By observing the three-axis attitude information of the sampling points, the polynomial is used to fit the target angle change. On this basis, the derivative can be used to obtain the function of speed and angular velocity with respect to time, and the intersection parameters such as the target speed, target attitude and flight trajectory can be solved.

[0029] D8: Use measured data to verify the feasibility and robustness of the measurement method. Use the control variable method to measure the effective ranging range of the lidar; use static tests to verify the accuracy of the method results under the condition of known target posture; use dynamic tests to verify the feasibility of the method in different environments and different motion states; finally, evaluate the size of the simulation test error and analyze the cause of the error.

[0030] The method and system for measuring and real-time solving of missile-target intersection parameters based on area array laser radar described in the present invention have the following effects: a tethered unmanned aerial vehicle carrying all measurement units of an observation end flies to a planned position for hovering observation; an area array laser radar and a visual sensor collect three-dimensional point cloud data and image data of a flight environment and a target at a fixed viewing angle and a fixed frame rate; a positioning and attitude determination system unit collects position and attitude data of the observation end; all data are transmitted in real time to an onboard industrial control computer unit for data storage and preprocessing; the unmanned aerial vehicle uses the preprocessed data to automatically avoid obstacles and adjust the observation angle; a wired communication unit transmits the preprocessed point cloud data, image data and position and attitude data of the observation end back to a display control unit and a missile-target intersection parameter solving unit at a ground workstation in real time; the missile-target intersection parameter solving unit solves the returned data in real time and transmits the solving result to the display control unit, thereby providing data support for the operator to perform supplementary measurements and adjust the observation plan.

[0031] Beneficial effects of the present invention:

[0032] The present invention collects the original point cloud data of the target in real time by observing the end-array laser radar, transmits it to the missile-target intersection parameter calculation unit at the ground workstation for real-time calculation, performs point cloud data processing operations such as point cloud extraction, point cloud denoising, and ground detection on the original data, and then uses algorithms such as axis extraction, ICP registration, and polynomial fitting to jointly calculate and display the missile-target intersection parameters such as the motion trajectory, intersection attitude angle, and position information of the flying target in real time. Compared with the prior art, the present invention has high measurement accuracy, strong real-time performance, strong anti-interference ability, and is suitable for multiple platforms and convenient to deploy, greatly improving the measurement accuracy and calculation efficiency of the missile-target intersection parameters. BRIEF DESCRIPTION OF THE DRAWINGS

[0033] Figure 1 It is a system structure diagram of the present invention;

[0034] Figure 2 The following is a working diagram of the observation end of the system of the present invention;

[0035] Figure 3 It is a flow chart of missile-target rendezvous parameter solution at the ground workstation of the system of the present invention;

[0036] Figure 4 1 is a diagram showing the effect of point cloud denoising in an embodiment of the present invention, wherein (a) is the original target point cloud, and (b) is a schematic diagram of the point cloud after denoising;

[0037] Figure 5 The ground detection result in the embodiment of the present invention;

[0038] Figure 6 : The target point cloud extraction result in an embodiment of the present invention, (a) is the point cloud projection, (b) is the point cloud extraction effect, and (c) is a schematic diagram of the point cloud after denoising. DETAILED DESCRIPTION

[0039] The specific implementation of the present invention is further described below in conjunction with the accompanying drawings and technical solutions.

[0040] The present invention provides a missile-target intersection parameter measurement and real-time solution system based on a planar laser radar. The system specifically comprises two parts: an observation end and a ground workstation end. The system structure of the present invention refers to Figure 1 The first part of the system of the present invention is the observation end. The work flow chart is as follows: Figure 2 The missile-target intersection parameter measurement system based on the area array laser radar provided by the present invention is used to measure the target missile-target intersection parameters. The specific measurement steps are as follows:

[0041] S100: The tethered drone is equipped with array laser radar, visual sensor, industrial computer, IMU, Beidou and other measurement equipment to fly to the pre-planned point for hovering observation. The array laser radar collects laser point cloud data of the target area at a fixed viewing angle and fixed frame rate, and the visual sensor collects image data of the target area.

[0042] S200: When the satellite positioning signal can be received, the Beidou module in the positioning and attitude system unit is used to obtain the position information (latitude, longitude and elevation) of the observation end, and the IMU module is used to obtain the attitude information of the observation end; when the satellite signal cannot be received or the satellite signal is weak, the IMU module and the array laser radar point cloud data are used to locate the UAV itself;

[0043] S300: The data collected in step S100 and step S200 are transmitted to the industrial computer unit in real time for data storage and preprocessing, so as to achieve rough stitching of point cloud data and provide data support for autonomous positioning and obstacle avoidance of the UAV;

[0044] S400: The industrial computer unit transmits the position and posture data collected in step S200 and the pre-processed point cloud data in step S300 back to the display control unit and the missile-target intersection parameter solving unit at the ground workstation end through the wired communication unit.

[0045] Step S200 specifically includes:

[0046] S201: The tethered UAV takes off from an open area, flies to the planned position and hovers. When the satellite signal is good, the Beidou module is used to obtain the UAV's position information, and the IMU module is used to collect the UAV's attitude information;

[0047] S202: When the satellite signal is weak (due to bad weather or tree obstruction), the point where the satellite is accurately positioned for the last time is used as the reference point, and the SLAM (Simultaneous Localization and Mapping) algorithm in the industrial computer is used to align the lidar data of adjacent frames. The position and attitude changes of the drone at two adjacent moments are estimated through the optimization algorithm, so as to cumulatively estimate the position and attitude changes of the drone relative to the reference point. At the same time, the accumulated error is corrected in combination with the IMU data, so as to achieve self-positioning of the drone.

[0048] Step S300 specifically includes:

[0049] S301: The collected observation system position and posture data, image data, and original point cloud data are stored in a data storage module in the industrial computer unit;

[0050] S302: The data preprocessing module in the industrial computer unit performs joint preprocessing on the collected data by removing point cloud outliers, reducing point cloud density, and enhancing point cloud features, thereby eliminating noise, reducing the amount of calculation, and improving the accuracy and robustness of subsequent algorithms, providing a higher quality data foundation for subsequent point cloud registration, segmentation, and reconstruction.

[0051] S303: First, invalid value elimination and radius filtering methods are used to filter outliers to reduce the impact of noise. Then, the voxel grid method is used to streamline the point cloud model and balance the distribution density of the point cloud data. The nearest neighbor difference method is used to perform feature sampling on the point cloud target to improve its resolution and enhance the point cloud features. Finally, the processed point cloud data is roughly aligned to generate a large-scale target point cloud model, which is then transmitted back to the ground workstation through the wired communication unit to provide a basis for the operator to perform supplementary measurements based on the roughly aligned point cloud information.

[0052] The second part of the system of the present invention is a ground workstation. To realize the real-time and interactivity of the system, the ground workstation is provided with a display control unit and a missile-target intersection parameter calculation unit, which provide the operator with real-time display and calculation of measurement data. The display control unit shares data with the missile-target intersection parameter calculation unit. The ground workstation display control unit includes the following steps:

[0053] S500: The industrial computer transmits the original data in the data storage module and the point cloud data processed by the data preprocessing module to the display control unit through the wired communication unit. The display control unit displays the position and posture of the observation end in real time, providing a basis for the UAV operator to control the UAV according to the surrounding environment, allowing the operator to adjust the position of the UAV and the parameter status of the onboard device in real time, and transmits the control instructions of the ground end to the observation end through the wired communication unit to complete supplementary measurement or manual obstacle avoidance operation of the UAV.

[0054] S600: The projectile-target intersection parameter solving unit shares data with the display control unit. The projectile-target intersection parameter solving unit transmits the solving result to the display control unit in real time to display the real-time solving result. The operator performs supplementary measurement or encrypted measurement on the target according to the solving result.

[0055] Reference for the calculation process of the missile-target intersection parameter calculation unit Figure 3 , which specifically includes the following steps:

[0056] S700: The measured data import and simulation data generation module in the missile-target intersection parameter calculation unit imports the position and posture data sent back from the industrial computer unit and the pre-processed point cloud data into the missile-target intersection parameter calculation unit. The point cloud image enhancement and filtering module performs operations such as denoising, target extraction, point cloud registration, and central axis extraction on the point cloud image. The target intersection parameter calculation module adopts the least squares method and polynomial fitting algorithms to jointly solve the missile-target intersection parameters such as target motion posture, target terminal speed and intersection posture angle, and transmits the solution results to the display control unit in real time.

[0057] S800: After the missile-target intersection parameter solution is completed, the measurement accuracy evaluation module evaluates the solution accuracy. The feasibility and robustness of the measurement method are verified using measured data, and the dual-mode fusion test is used to verify the effective ranging range of the laser radar using the control variable method under the existing test equipment; static tests are used to verify the accuracy of the method results under the condition of known target posture; dynamic tests are used to verify the feasibility of the method in different environments and different motion states.

[0058] Step S700 specifically includes:

[0059] S701: Based on the original point cloud data, the k-means clustering algorithm is used to perform point cloud denoising. For complex and irregular point cloud models, the k-means clustering algorithm has a good clustering effect, and its algorithm time complexity is relatively low. According to the data situation, first determine the initial cluster center of the data and the calculated k value, then calculate the distance from the point to the initial cluster center, determine the distance size, summarize to form a new cluster, and continue clustering until the cluster center no longer changes. The Euclidean distance d(x,C i )Calculation formula:

[0060]

[0061] In the formula, x is the data object; C i is the cluster center; m is the dimension; x j , C ij For x and C i The attribute value of .

[0062] The formula for calculating the sum of squared errors (SSE) for the entire data set is:

[0063]

[0064] For complex and irregular point cloud models, the k-means clustering algorithm has a good clustering effect, and its algorithm time complexity is relatively low. Figure 4 .

[0065] S702: Ground detection based on plane fitting. The present invention uses the RANSAC algorithm to perform ground recognition. The points on the plane satisfy the following formula:

[0066] ax+by+cz+d=0

[0067] In the formula, d is the distance from the plane to the origin; a, b, and c are the Cartesian components of the plane normal vector. It can be seen from the laser radar point cloud data that most point clouds scan the ground, so the point cloud data is used to fit this plane model, and a largest plane is found so that as many point clouds as possible satisfy this model. At this time, this plane is the ground plane. Due to the slight undulation of the ground locally and the detection accuracy of the array laser radar, all ground points do not belong to the ground plane model exactly, so it is necessary to relax the model constraints, that is, when the distance from the point to the plane model does not exceed the threshold, the point is considered to belong to the ground point. In addition, since the surface of tall buildings in the scene may affect the parameters of the optimal plane model, when fitting the plane model parameters, its angle relative to the horizontal plane is calculated. If the angle is greater than the threshold, the model is considered not to be a ground model. At the same time, considering that the ground may have a horizontal plane and a slope plane at the same time, the RANSAC algorithm can only identify one of the road surfaces. After the RANSAC algorithm is executed, the RANSAC algorithm is continued to be executed on the outliers to achieve two ground recognitions and identify the horizontal plane and the slope plane.

[0068] The RANSAC algorithm is used to identify the ground point cloud in the lidar point cloud data. The recognition results are shown in Figure 5 .Depend on Figure 5 It can be seen that the RANSAC algorithm can be used for plane estimation to accurately detect the ground from the point cloud data. Retaining all point cloud data other than the ground can eliminate the interference of the ground point cloud during clustering.

[0069] S703: After identifying the ground point cloud, the point cloud target is extracted by dual-modal fusion of image and point cloud. First, the improved background difference method is used to realize image target extraction, as follows: on the basis of the target detection algorithm based on background difference and inter-frame difference, the judgment of the surrounding points of adjacent frames is added to better remove the foreground noise and extract the pixel information of the target object. This method uses a fixed threshold when extracting image targets. There is a situation where the background cannot be completely removed. For the background noise that has not been removed, the maximum contour extraction algorithm based on Opencv is used to complete the maximum contour extraction. Finally, the area of ​​the target projection point cloud is determined by comparing the target area and the projection point cloud coordinates.

[0070] Secondly, extract the point cloud, as follows: After the point cloud is projected to the image pixel coordinate system, the relative position of the point cloud does not change. Therefore, by observing the position of the target projection point cloud in the overall projection point cloud, the position of the target 3D point cloud in the original point cloud can be obtained. When selecting the point cloud in the target projection point cloud range, it is easy to extract background point cloud information at the edge. Therefore, the extracted point cloud target is finally denoised using the point cloud denoising algorithm in step S701 to obtain the target 3D point cloud. The target extraction effect is shown in Figure 1. Figure 6 .

[0071] S704: Based on the extracted 3D point cloud data, the iterative closest point (ICP) algorithm is used to complete the registration of the target point cloud and the point cloud model, and obtain the pose transformation matrix. The rough registration of the point cloud data by the observation end data preprocessing module provides a good initial position for the fine registration. The ICP algorithm continuously iterates on this basis, and performs rotation and translation through the registration between points, further reducing the error to infinitely approach the ideal accuracy value.

[0072] S705: After ICP registration is completed, the 4×4 rotation and translation matrix from the point cloud model to the target point cloud can be obtained, but the rotation matrix of the target point cloud with respect to the body-centered coordinate system cannot be directly obtained through the rotation and translation matrix. In order to obtain the rotation matrix and spatial position of the target point cloud with respect to the body-centered coordinate system, it is necessary to first perform normal fitting based on the rotated point cloud model, and then use the least squares method to fit the target surface to achieve the axis. For situations where the number of point clouds is small or the point clouds are incomplete, it is first necessary to complete the point cloud data, use the voxel grid algorithm to convert the point cloud data into a three-dimensional network, and fill the missing areas through interpolation to achieve the purpose of point cloud completion.

[0073] S706: Use a polynomial to fit the motion trajectory of the target. The polynomial is expressed as:

[0074]

[0075] Where t is time, P(t) is the fitting position value at time t, and a i is the coefficient, n is the order of the fitting polynomial, and m is the number of sampling points. When n+1<m, an overdetermined system of equations can be constructed. Finding the least squares solution to the overdetermined system of equations can suppress random errors in the data and achieve the purpose of data smoothing. The order of the fitting polynomial n can generally be 3 or 4. According to the target point cloud data, the time, position, angle and other information of the moving target can be obtained, so the motion trajectory can be abstracted as a polynomial of x, y, z with respect to time.

[0076] S707: The test background of the embodiment of the present invention is low-altitude observation (observation height is 80 meters), and the projectile throwing direction is fixed, so the flip angle along the xOy and xOz planes is ≤360°. Therefore, the attitude change is estimated by the curve fitting method. According to step S704, the three-axis attitude information can be solved, and the angle change of the target can be fitted by a polynomial. The angle fitting polynomial is: p=a0+a1t+a2t 2 +a3t 3 By taking the derivative of the angle fitting polynomial and the displacement variation polynomial fitted in step S706, the function of the target speed and angular speed with respect to time can be obtained, and then the target speed and target coordinates can be solved according to the height of the target from the ground and the flight time when it is dropped.

[0077] After the real-time solution of the missile-target rendezvous parameters is completed, it is necessary to conduct an accuracy evaluation of the solution parameter results. Through simulation tests and ground verification tests, the feasibility of the pose estimation scheme and the accuracy of the pose estimation algorithm are verified respectively.

[0078] Step S800 specifically includes:

[0079] S801: Divide the distance between the target and the LiDAR into multiple groups, conduct a point cloud and image dual-modal fusion test, verify the fusion effect of point cloud and image data according to different object distances, and use the control variable method to measure the effective ranging range of the LiDAR.

[0080] S802: According to the test conclusion of step S801, under the condition of known target posture, a static test is performed to verify the accuracy of the algorithm result described in S700. The embodiment of the present invention selects a distance between the laser radar and the target of 9 to 12 meters to complete the static test to verify the error of the algorithm at different object distances.

[0081] S802: In different environments and different motion states, dynamic tests are used to verify the feasibility of the algorithm in step S700. The target is thrown under different environmental conditions and at different object distances. The test data is obtained, and the accuracy of the measurement results is evaluated by manually achieving the specified target position.

[0082] The above-mentioned test was used to evaluate the accuracy of the measurement results, and the accuracy and stability of the method were verified using simulation tests and semi-physical ground tests. The results showed that the accuracy of the two-dimensional image target detection algorithm was 97%, and the misclassification ratio was 0.0112%; the accuracy of the pose estimation algorithm was improved by 53.2% compared with the traditional algorithm, and the single-shot time was reduced from 261ms to 132ms, with an efficiency improvement of about 49.4%; the angle error of the landing point angle obtained by the polynomial fitting method was between 0° and 0.003° compared with the original angle.

Claims

1. A missile-target intersection parameter measurement and real-time solution system based on array laser radar, characterized in that: The system includes an observation end and a ground workstation end. The observation end includes a planar laser radar unit, a visual sensor unit, a positioning and attitude system unit, an industrial computer unit, a wired communication unit and a power supply transfer module. The ground workstation end includes a display control unit, a missile-target intersection parameter calculation unit, a wired communication unit and a tethered power supply unit. The array laser radar unit is connected to the industrial computer unit, the positioning and attitude determination system unit, the wired communication unit and the power adapter module to collect true color three-dimensional point cloud data of the target movement in real time, and transmit the data to the industrial computer in real time for storage, forwarding and data preprocessing; The visual sensor unit is connected to the industrial computer unit and the power adapter module, takes pictures of the target area at a certain frame rate, and transmits the collected image data to the industrial computer for storage, forwarding and image data preprocessing; The positioning and attitude determination system unit is composed of an IMU module and a Beidou positioning module. The IMU module is used to obtain the speed and attitude information of the observation end, and the Beidou positioning module is used to obtain the latitude, longitude and altitude information of the observation end. The collected information is transmitted to the industrial control computer unit in real time for data storage and forwarding; The industrial computer unit is composed of a data storage module and a data preprocessing module, and the area array laser radar unit, the visual sensor unit and the positioning and attitude system unit are all connected to the industrial computer unit; the data storage module is used to store the original point cloud data of the target movement collected by the area array laser radar unit, the target area image data collected by the visual sensor, and the observation system position and attitude data collected by the positioning and attitude system unit, and the data preprocessing module is used to jointly solve the original point cloud data, image data and position and attitude data to achieve rough splicing of the point cloud data and generate a large-scale three-dimensional environment model; The wired communication unit includes a wired communication unit at the observation end and a wired communication unit at the ground end. The two wired communication units transmit data through a network cable and are responsible for communication transmission services at the observation end and the ground end respectively. The wired communication unit at the observation end is connected to the industrial computer, and transmits the position and posture data collected by the positioning and attitude determination system unit, the image data collected by the visual sensor unit, and the point cloud data pre-processed by the industrial computer unit back to the ground end in real time. The wired communication unit at the ground end transmits the data to the display control unit and the missile-target intersection parameter solution unit in real time, and transmits the observation command issued by the operator back to the observation end. The power adapter module is connected to the tethered power supply unit at the ground end through a tethered cable to provide safe and stable power supply support for the area array laser radar unit, visual sensor unit, industrial computer unit and positioning and attitude determination system unit at the observation end; The display control unit shares data with the missile-target intersection parameter solving unit, and displays two parts of data information transmitted by the observation end and the missile-target intersection parameter solving unit on the ground end, one part of which displays the pre-processed point cloud data and the position, posture and other information of the observation system transmitted back in real time by the observation end, and the other part displays the parameter settlement result transmitted back by the missile-target intersection parameter solving unit, so as to provide decision information for the operator to perform supplementary measurement and encrypted measurement; The tethered power supply unit is connected to the power adapter module through a tethered cable to provide safe and stable power supply support for the observation end and the ground end; The missile-target intersection parameter calculation unit is composed of a measured data import and simulation data generation module, a point cloud image enhancement and filtering module, a target intersection parameter calculation module and a measurement accuracy evaluation module, and is connected to a data wired communication unit and a display control unit; the measured data import and simulation data generation module imports the measured data collected by the array laser radar in real time, and uses the STL model of the typical scene model library imported from VTK, mainly including airports, ports, and land buildings; the point cloud image enhancement and filtering module uses a dual-modal information fusion algorithm and a RANSAC algorithm to perform point cloud target extraction and ground detection, and uses a K-means clustering algorithm to perform point cloud denoising; The target rendezvous parameter solution module estimates the target's motion trajectory through polynomial fitting based on the extracted target information and axis equation, and solves the rendezvous parameters such as motion posture, target terminal velocity and rendezvous posture angle; the measurement accuracy evaluation module uses the average Euclidean distance to define the error result between the solved target information and the simulation information, and realizes comparative analysis and accuracy evaluation functions by comparing the solved target information with the simulation information.

2. A method for measuring and real-time solving missile-target intersection parameters based on array laser radar, characterized in that: The system of claim 1 comprises the following steps: Step A: All measurement units at the observation end are loaded onto a tethered drone, and the drone is flown to a planned location for hovering observation; the array laser radar and visual sensor collect the target three-dimensional point cloud data and target area image data in real time at a fixed frame rate and viewing angle, and the positioning and attitude determination system unit collects the position and attitude data of the observation end. The collected data are all transmitted to the industrial computer unit for data storage, forwarding and preprocessing; Step B: The data preprocessing module in the industrial computer unit jointly solves the collected observation end position and attitude data, image data and original point cloud data, completes the rough stitching of the point cloud data, and solves the observation end position and attitude changes to achieve self-positioning and obstacle avoidance navigation of the UAV; Step C: The wired communication unit transmits the rough splicing point cloud data and the position and attitude data of the observation end preprocessed by the industrial control computer to the display control unit of the ground workstation in real time. The display control unit transmits the data to the missile-target intersection parameter solving unit for real-time solving, and sends the solving result to the display control unit. The operator controls the position of the tethered UAV and the direction of the adjustable self-stabilizing gimbal on the aircraft according to the real-time solving result, and performs supplementary measurement on the key targets. Step D: The missile-target intersection parameter calculation unit performs real-time calculation on the returned data, including point cloud denoising, target extraction, registration, and trajectory fitting. After the calculation is completed, the calculation result is sent back to the display control unit, as follows: D1: K-means clustering algorithm is used to set the radius threshold, compare the Euclidean distance of randomly selected points, and delete all points larger than the radius threshold to achieve denoising, so as to eliminate noise points other than the target; D2: Identify and remove the ground point cloud, and only keep the obstacle point cloud; then use the RANSAC algorithm to fit the plane, and divide the original data into two parts: internal points and external points. The internal points satisfy the plane model and belong to the ground points, while the external points do not satisfy this model and belong to obstacle points. D3: A moving target extraction method based on dual-modal information fusion is adopted to establish a coordinate transformation model between the color camera and the array laser radar to achieve pixel-level matching of the two sensors, and to fuse the image and point cloud at the same time. The details are as follows: First, an improved background difference algorithm is used to extract the moving target in the image, and the mean filter is used to remove the image noise. The algorithm in the OpenCV library is used to complete the image target contour extraction. Finally, the corresponding point cloud is selected according to the position of the image target. Then, after the point cloud is projected to the image pixel coordinate system, the position of the target three-dimensional point cloud in the original point cloud can be obtained by observing the position of the target projection point cloud in the overall projection point cloud. D4: After completing the segmentation of the target point cloud data, the ICP algorithm is used to realize the registration of the target point cloud and the point cloud model, obtain the posture conversion matrix, and estimate the posture information of the missile; D5: Based on the rotated point cloud model, normal fitting is performed first, and the cylindrical surface fitting of the point cloud target is performed using the least squares method to extract the axis; D6: Based on the extracted target information and axis equation, the polynomial fitting method is used to estimate the target's actual motion trajectory. The trajectory prediction method based on Gaussian mixture model is used to model complex motion patterns, calculate the probability distribution of different motion patterns, divide the trajectory data into different components, and use Gaussian process regression to predict the most likely motion posture of the moving object. D7: The time, position and angle information of the target are obtained based on the point cloud data collected by the array laser radar. The target angle change is abstracted as a polynomial of the target's rotation angle in the xOy and xOz planes with respect to time. By observing the three-axis attitude information of the sampling points, the polynomial is used to fit the target angle change, and the velocity and angular velocity functions with respect to time are derived to solve the intersection parameters such as the target speed, target attitude and flight trajectory. D8: Use measured data to verify the feasibility and robustness of the measurement method; use the control variable method to measure the effective ranging range of the lidar; use static tests under the condition of known target posture to verify the accuracy of the method results; use dynamic tests in different environments and different motion states to verify the feasibility of the method; finally, evaluate the size of the simulation test error and analyze the cause of the error.

3. The method for measuring and calculating missile-target intersection parameters based on area array laser radar according to claim 2 is characterized in that: The improved background difference method is as follows: on the basis of the target detection algorithm based on background difference and inter-frame difference, the judgment of the surrounding points of adjacent frames is added to extract the pixel information of the target object; then the maximum contour extraction algorithm based on Opencv is used to complete the maximum contour extraction; finally, the area of ​​the target projection point cloud is determined by comparing the target area and the projection point cloud coordinates.

Images