Space moving target optical comb rapid distance measurement method and system based on hierarchical guidance strategy

Abstract

The invention discloses a spatial moving target optical comb rapid distance measurement method and system based on a hierarchical guidance strategy, relates to the technical field of spatial moving target distance measurement, and aims to solve the problems of long alignment time, high energy consumption and high efficiency caused by the fact that traditional optical alignment before moving target scanning depends on a preset global scanning mode. And the requirements of dynamic tracking distance measurement are difficult to meet. The system is divided into a laser emitting and scanning part, an image receiving and processing part and a processor part. The laser emission scanning part utilizes the deflection of a fast reflecting mirror to change a laser light path, and scans a reflecting prism on the surface of a target; the image receiving and processing part reflects an image of a target reflecting prism to a binocular camera through a small fast reflecting mirror, and then inputs an imaging result of the camera into an image processing module to calculate a relative pose with a target; and the main processor performs target motion estimation and scanning path planning according to the relative pose information, and inputs reference path information to the fast steering mirror driving circuit to implement tracking control. Through a multi-sensor fusion sensing algorithm, accurate target sensing and positioning in a complex space environment are realized, interference such as in-orbit platform vibration and space illumination condition change under a microgravity condition can be effectively coped with, and thus the robustness and accuracy of space target detection are greatly improved.

Description

Technical Field

[0001] The present invention relates to the technical field of space moving target ranging, and in particular to a method and system for fast ranging of space moving targets using an optical comb based on a layered guidance strategy. Background Art

[0002] With the rapid development of my country's aerospace technology, the increasing number of space missions, such as spacecraft rendezvous and docking, satellite cluster formations, precision measurement of large space structures, and satellite antenna positioning, has placed higher demands on precise distance measurement capabilities in space. The femtosecond laser optical frequency comb (or optical comb for short) is a new type of broadband coherent light source. Its time domain characteristics manifest as a series of femtosecond ultrashort pulses, while its frequency domain characteristics present a uniformly distributed comb-like spectrum. It has the advantages of high stability, narrow pulse width, and high bandwidth. It has opened up a new technical path for high-precision absolute distance measurement and has attracted widespread attention in the field of high-precision distance measurement in space, both domestically and internationally. In 2013, South Korea experimentally deployed an erbium-doped fiber comb light source to a low-Earth satellite, verifying its ability to work in emission vibration, high-energy radiation and thermal vacuum environments; in 2021, the FOKUS II dual-comb light source system developed by Germany conducted a space payload test, verifying its in-orbit self-starting capability; in 2022, the low-noise fiber comb light source developed by the Chinese Academy of Sciences, as the core component of the high-precision time and frequency experiment cabinet, was successfully carried on the Tiangong space station and carried out in-orbit tests, fully verifying its ability to provide high-precision time and frequency signals in a space environment.

[0003] Femtosecond laser comb ranging technology uses a comb laser as its light source. By linearly sampling two combs with a certain repetition frequency difference, it can simultaneously obtain large-scale time-of-flight information in the time domain and high-precision interference information in the frequency domain, thereby achieving high-precision and high-speed absolute distance measurement. In 2018, the Karlsruhe Institute of Technology in Germany successfully achieved precise tracking of high-speed moving objects through synthetic wavelength interferometry using two microcavity combs. In 2021, Tsinghua University combined dual-comb ranging with spectral phase resolution technology to simultaneously measure the absolute distance and attitude angle information of the target, achieving a ranging accuracy of 13.7 nanometers and an angular accuracy of 0.43 microradians, demonstrating great application potential in the field of high-precision distance and posture measurement in space.

[0004] The alignment speed and accuracy of the laser and the target reflective prism directly affect the time efficiency and dynamic stability of optical comb ranging. In ground-based optical comb ranging experiments, the reflective prism and laser of the target are usually pre-aligned in a fixed mode to improve the accuracy of static measurements. However, in space applications, the ranging platform and the target are often in relative motion, making it difficult to maintain a stable alignment. The current alignment method requires a preset multi-mode scanning strategy to periodically and globally cover the plane where the target reflective prism is located. However, this unguided blind scanning mechanism has the problems of long scanning cycles and easy target omissions, making it difficult to meet the real-time tracking and ranging requirements of dynamic targets. Therefore, how to effectively utilize the sensors carried by the on-orbit platform, guide the optimization of the scanning path and compress the search space through active sensing technology, and build a high-efficiency and high-precision dynamic scanning strategy is a technical problem that needs to be overcome. Summary of the Invention

[0005] The technical problems to be solved by the present invention are:

[0006] This paper addresses the challenges presented by the prior art by proposing a rapid laser scanning path planning method based on a hierarchical guidance strategy to achieve rapid alignment and tracking of dynamic targets. This algorithmic innovation aims to support the application of femtosecond optical comb ranging technology in the field of on-orbit precision ranging. This addresses the issues of traditional optical alignment for moving targets, which relies on a preset global scanning pattern before scanning, resulting in lengthy alignment times, high energy consumption, and difficulty meeting the requirements of dynamic tracking and ranging.

[0007] The technical solution adopted by the present invention to solve the above technical problems is:

[0008] The system for rapid optical comb ranging of a moving target in space based on a layered guidance strategy, according to the present invention, comprises a laser emitting and scanning section, an image receiving and processing section, and a processor section. The laser emitting and scanning section comprises a laser emitter, a large quick-reflection mirror, and a large quick-reflection mirror driving circuit. The image receiving and processing section comprises a small quick-reflection mirror, a binocular camera, an image processing module, and a small quick-reflection mirror driving circuit. The laser emitter is used to emit a laser beam for optical comb ranging. The deflection of the large quick-reflection mirror is used to change the direction of the laser light path, so that the laser scans a target platform and ultimately aligns with a reflective prism mounted thereon. The large quick-reflection mirror driving circuit is used for bottom-level control, driving the quick-reflection mirror to accurately track a reference path provided by a main processor. The image of the target reflective prism is reflected to the binocular camera via the small quick-reflection mirror, and the imaging result of the binocular camera is then input into the image processing module. The small quick-reflection mirror is used to reflect external light to the binocular camera. The small quick-reflection mirror does not undergo large-angle deflection and is only used for fine-tuning and calibrating the reflected light path. The image processing module is used to de-jitter, extract features, and fuse images captured by the binocular camera, and to calculate the relative position and posture of the target. The main processor performs target motion estimation and scanning path planning based on the relative posture information, and inputs the reference path information into the fast mirror drive circuit to implement tracking control.

[0009] Furthermore, the binocular camera is a D435i binocular camera; the main processor uses the NVIDIA Jetson Xavier NX high-performance edge computing module, which receives relative pose information sent by the image processing module and executes the trajectory prediction and path planning algorithm. A small quick-reflector mirror reflects external light through a beam splitter prism to the binocular camera and the optical comb ranging module. The image reception and processing section also includes a lens assembly positioned between the beam splitter prism and the binocular camera. The small quick-reflector mirror reflects external light through the beam splitter prism, and then the laser enters the binocular camera through the lens assembly. The lens assembly is used to improve optical quality, control exposure, and reduce glare.

[0010] The laser emission and scanning section uses the deflection of a quick-reflection mirror to change the laser light path and scan the reflective prism on the target surface. The image receiving and processing section reflects the image of the target reflective prism to the binocular camera through a small quick-reflection mirror. The camera imaging result is then input into the image processing module to resolve the relative position with the target. The main processor estimates the target motion and plans the scanning path based on the relative position information, and inputs the reference path information into the quick-reflection mirror drive circuit to implement tracking control. The method for rapid ranging of spatial moving targets with optical combs based on a hierarchical guidance strategy is implemented in the following steps:

[0011] Step 1: Acquisition of the relative pose of the target based on multi-sensor fusion. Given that on-orbit platforms are often affected by vibration and microgravity, which may lead to deterioration of image quality, an image de-shaking algorithm is introduced. In order to accurately estimate inter-frame motion, a Kalman filter and vision-IMU fusion technology are used, and the pose output is optimized by analyzing the covariance matrix, thereby suppressing drift errors and ensuring the accuracy and stability of image information. In addition, since the quality of RGB images in a space environment is extremely susceptible to changes in light, a D435i camera is used to simultaneously capture RGB and infrared images of the target, thereby providing additional visual information when lighting conditions are poor. Subsequently, this method uses the SIFT algorithm to efficiently extract feature points from these two images, and improves the recognition accuracy and robustness of image content through feature matching.

[0012] On this basis, the relative position of the camera and the target reflective prism is accurately calculated using the debounced image and matched feature points using the PnP algorithm. Furthermore, by combining the coordinate transformation matrix between the large quick-reflection mirror in the laser emission module and the camera, the relative position between the quick-reflection mirror and the reflective prism can be calculated. Finally, to apply this relative position information in actual control, it is projected into the two-dimensional space of the XY-axis mechanical scanning angle of the large quick-reflection mirror drive mechanism, providing a data foundation for planning the quick-reflection mirror's deflection path.

[0013] Step 2: Relative pose information prediction based on the recursive convolutional conditional neural process algorithm. The conditional neural process (CNP) is a type of deep neural network that describes a Bayesian Gaussian process and can predict the distribution of an unknown sequence through a contextual data sequence. First, a convolutional neural network (CNN) is used as the encoder, and the historical pose sequence near the current moment is used as the encoder input. The local temporal information is extracted through a one-dimensional convolutional network and mapped into a latent vector h. Subsequently, a multi-layer perceptron (MLP) is used as the decoder, and the latent vector h and the timestamp of the pose to be queried are used as input. Finally, the Gaussian distribution of the relative pose of the reflecting prism and the large fast mirror at the future moment is output.

[0014] In order to further improve the model's temporal feature modeling capabilities, a recurrent neural network (RNN) architecture is introduced into the outer layer of the CNP network, allowing the CNP to recursively predict the pose information of multiple steps in the future. At the same time, the negative log-likelihood (NLL) is used as the loss function to ensure that the predicted distribution can maximize the likelihood of the real data, thereby improving the accuracy of the prediction.

[0015] Then, based on the predicted position and distribution information of the trajectory points, the potential field distribution map within the workspace is calculated by performing a time-decayed weighted summation of the potential field generated for each predicted point and adding the repulsive field relative to the obstacle. When calculating the potential field of each predicted point, it is necessary to consider the Euclidean distance within the potential field space and the probability density distribution of the predicted point itself. At the same time, by setting a time exponential decay weight for each potential field, predicted points closer to the current time receive a greater weight, thereby generating a time-dependent potential field map. In addition, considering that the target reflective prism has an incident angle dead zone, there may be areas in the workspace that cannot be effectively scanned. We set higher potential field values ​​for these areas to avoid energy loss caused by ineffective scanning.

[0016] Step 3: B-spline coverage path dynamic planning algorithm based on particle swarm optimization. The artificial potential field map obtained in step 2 reflects the probability of the target reflective prism appearing in the workspace. In order to improve the scanning efficiency, we only scan the areas where the potential field is below a certain threshold. However, due to the complexity and diversity of the potential field changes, these scanning areas below the potential field threshold often have irregular shapes. In order to simplify the complexity of coverage path planning, we first decompose the scanning area into multiple simple and non-overlapping work unit areas with different prediction points as the center. These areas constitute the workspace for coverage path planning. The work units are divided according to the potential field threshold equipotential lines near each prediction point. However, due to the significant potential field changes between different prediction points, in order to ensure that the path planning has sufficient workspace, the division of different work units refers to different equipotential lines.

[0017] The coverage path is constructed using spirals drawn using B-spline curves. The planning process consists of two steps: pre-planning and online optimization. The main goal of pre-planning is to construct evenly spaced irregular spirals that cover the workspace. The control points of the B-spline are represented in polar coordinates, and the required spirals are generated by adjusting the length coordinate spacing of control points with the same arc. During online optimization, the key metrics for the optimal coverage path are spiral density, coverage area, and path length. The density of the spirals ensures more detailed scanning of areas with a high probability of target occurrence; the coverage area of ​​the spirals should cover the entire work cell as much as possible; and the path length determines the scanning time. An excessively long path may result in the scanning being incomplete before the target reflector moves out of the work cell. The starting point for online optimization is the spiral control points generated in pre-planning. Using a particle swarm optimization algorithm, the polar coordinates of each control point and the number of spiral turns are fine-tuned. A loss function is constructed based on the spiral density, path length, and coverage area to ultimately optimize the optimal coverage path within each work cell.

[0018] However, the neural network-based trajectory prediction in step 2 cannot be updated in real time. Therefore, before the prediction is updated, the original potential field must be updated based on the target pose updated by the vision module, and the work units and planned path must be readjusted. If the original predicted path is relatively accurate, the new path plan can simply involve fine-tuning the old B-spline control points. Ultimately, the optimized optimal path is sent to the fast-reflection mirror driver module for scanning, as the tracking target.

[0019] The present invention has the following beneficial technical effects:

[0020] This invention effectively solves the time-consuming and energy-consuming problem of the optical alignment scanning process in traditional optical comb ranging. The invention proposes a high-efficiency and fast laser scanning coverage path planning algorithm based on a layered guidance strategy, and fast ranging of space moving targets with optical combs (lasers). The optical comb ranging for space moving targets proposed by the invention adopts a layered guidance strategy, which can achieve efficient and fast laser scanning path planning. The beneficial effects of the invention are as follows:

[0021] 1. Through multi-sensor fusion perception algorithms, accurate target perception and positioning in complex space environments are achieved. It can effectively cope with interference such as on-orbit platform vibration under microgravity conditions and changes in space lighting conditions, thereby greatly improving the robustness and accuracy of space target detection.

[0022] 2. Using a recursive convolutional conditional neural process algorithm, we predict the position of the target reflective prism over a period of time. By introducing convolutional layers and a recursive neural network architecture, we further enhance the neural network's ability to extract temporal features from the data. Using negative log-likelihood as the loss function, we achieve accurate and reliable predictions over large timescales based on small sample training data.

[0023] 3. Accurate and efficient scanning path generation is achieved through a layered overlay path planning algorithm. The artificial potential field-based mapping comprehensively considers the relative distance of the prediction points, probability density distribution, time decay, and laser scanning blind spots, accurately describing the possible locations of the target reflective prism. The introduction of a work unit segmentation method simplifies the complexity of planning within complex workspaces and avoids repetition of planned paths. B-spline curves are used to construct the path, and pre-planning and particle swarm optimization algorithms are used to adjust control points, improving planning efficiency and effectiveness. BRIEF DESCRIPTION OF THE DRAWINGS

[0024] Figure 1 It is the system device and operation structure diagram;

[0025] Figure 2 This is the structural block diagram of the image processing module;

[0026] Figure 3 This is the multi-sensor fusion positioning effect diagram;

[0027] Figure 4 This is the structure diagram of the recursive convolutional conditional neural network;

[0028] Figure 5 This is the training effect diagram of recursive convolutional conditional neural network;

[0029] Figure 6 This is the predicted trajectory distribution effect diagram of the recursive convolution conditional neural process;

[0030] Figure 7 This is the structural diagram of the hierarchical guided planning algorithm;

[0031] Figure 8 The structure of the reflecting prism (corner cube prism) and the reflected light path diagram;

[0032] Figure 9 This is the artificial potential field effect diagram generated based on the trajectory prediction distribution;

[0033] Figure 10 This is the effect diagram of the hierarchical coverage path planning algorithm. DETAILED DESCRIPTION

[0034] Combined with attachment Figure 1-10 The present invention specifically describes an efficient and rapid coverage path planning system and method based on a hierarchical guidance strategy for space laser rapid ranging.

[0035] The efficient and fast coverage path planning system based on hierarchical guidance strategy for space laser rapid ranging described in the present invention is as follows: Figure 1The figure shows the device and operational block diagram of the system according to the present invention. The entire system is divided into a laser emission and scanning section, an image reception and processing section, and a processor section. The laser emission and scanning section consists of a laser emitter, a large quick-reflection mirror, and a large quick-reflection mirror drive circuit. The laser emitter is used to emit the laser beam used for optical comb ranging. The large quick-reflection mirror changes the direction of the laser light path through its own deflection, allowing the laser to scan the target platform and ultimately align with the reflective prism mounted on it. The large quick-reflection mirror drive circuit provides low-level control, driving the quick-reflection mirror to accurately track the reference path provided by the main processor. The image reception and processing section consists of a small quick-reflection mirror, a lens assembly, a D435i binocular camera, an image processing module, and a small quick-reflection mirror drive circuit. The small quick-reflection mirror reflects external light to the binocular camera and optical comb ranging module. It does not perform large-angle deflection and is only used for fine-tuning and calibration of the reflected light path. The lens assembly improves optical quality, controls exposure, and reduces glare. The image processing module is used to eliminate jitter, extract and fuse features, and calculate relative pose of images captured by the binocular camera. The main processor uses NVIDIA's Jetson Xavier NX high-performance edge computing module to receive relative posture information sent by the image processing module and execute related algorithm programs such as trajectory prediction and path planning.

[0036] Let's explain the function of the large quick reflex mirror: the basic function of the large quick reflex mirror is to deflect the laser so that it scans the target surface. When the laser scans the reflective prism on the target surface, it will be reflected back to the small quick reflex mirror, and then reflected by the small quick reflex mirror to the optical comb ranging module. When the optical comb ranging module detects the reflected laser, it will send an instruction to the onboard computer. After receiving the instruction, the onboard computer will control the driving circuit of the large quick reflex mirror to stop scanning, so that the laser remains aligned with the reflective prism on the target platform. However, in reality, it is impossible for two spacecraft platforms in orbit to remain relatively still. Therefore, the guidance algorithm (method part) described in the present invention is needed to assist the large quick reflex mirror in quickly aligning the reflective prism and shorten the time interval between the two alignments. The main purpose of the system provided by the present invention is to shorten the time it takes for the large quick reflex mirror to deflect the laser to align with the reflective prism.

[0037] The optical comb ranging module can be implemented using existing technology. The optical comb ranging module is described in this invention for the sake of completeness of the system description. The laser reflected by the target reflection prism is reflected by a small fast mirror and split into two paths by a beam splitter prism. One path enters the binocular camera through the lens group, and the other path enters the optical comb ranging module. Figure 1 The main purpose of the system provided by the present invention is to shorten the time it takes for a large quick-reflection mirror to deflect laser light toward a reflecting prism.

[0038] The implementation process of the efficient and rapid coverage path planning method based on a hierarchical guidance strategy for space laser rapid ranging described in the present invention is as follows:

[0039] Step 1: Accurate target perception and positioning in complex spatial environments based on multi-sensor fusion perception algorithms.

[0040] like Figure 2 As shown in the figure, the equipment and flow chart of the image processing module are displayed. First, the IMU and camera information are fused to compensate for the image jitter caused by the vibration of the on-orbit platform. Then, the infrared camera and RGB camera are fused to adapt to the changes in spatial illumination. Finally, the relative pose is solved according to the coordinates of the feature points.

[0041] The IMU mounted on the camera platform can measure the platform's acceleration and angular velocity. The image's motion information can also be described by x-axis translation pixels dx, y-axis translation pixels dy, and rotation angle da. By fusing the high-frequency components of these two motion signals through a Kalman filter, the platform's vibration estimate can be obtained. Let T be the fused state estimate, and the fusion process can be expressed as:

[0042] T k|k =T k|k-1 +K k (z k -H·T k|k-1 )

[0043] where x k|k-1 is the predicted state, K k is the Kalman gain, z k is the observation value, and H is the observation matrix. Subsequently, the image pixels captured by the camera can be compensated based on the vibration estimate to eliminate the image distortion caused by the platform vibration:

[0044] T compensated =TT motion

[0045] The SIFT algorithm is a widely used algorithm for image feature extraction and feature matching. It can extract stable local feature points at different scales and rotation angles. We use the image information collected by the D435i binocular camera and perform feature extraction using the SIFT algorithm to obtain feature point descriptors in RGB and infrared images, which are expressed as:

[0046] D RGB =SIFT(Image RGB ),D IR =SIFT(Image IR )

[0047] Then, by performing feature-level fusion on the feature points of the two, the joint feature vector set D is obtained. fusion =[D RGB ,D IR]. This feature set integrates all feature information collected by the RGB camera and the infrared camera, which can significantly improve the accuracy and stability of target recognition under conditions of spatial illumination changes.

[0048] The PnP algorithm can estimate the camera's pose from several known three-dimensional coordinate points and their two-dimensional projections on the image. First, obtain the three-dimensional coordinates P of the preset calibration point on the target platform. i And obtain the two-dimensional coordinate p through the camera i , and then solve the camera's external parameters, that is, the rotation matrix R, through direct linear transformation, EpnP or Lvenberg-Marquardt method c and T c , three-dimensional coordinates P i To the two-dimensional coordinate p i The projection relationship is:

[0049] p i =K·[R c |T c ]·P i

[0050] Among them, K is the known intrinsic parameter matrix of the camera, which mainly contains information such as the focal length and principal point of the camera.

[0051] Figure 3 The positioning accuracy of RGB cameras, infrared cameras, and IMU visual fusion was compared. It can be seen that the inherent platform vibration causes a certain amount of high-frequency noise in the positioning results of both cameras. Furthermore, due to the susceptibility of RGB and infrared cameras to light and ambient temperature, the independent positioning results all exhibit errors. By combining IMU-visual de-jitter and RGB-IR image feature fusion, the accuracy and stability of visual positioning are improved.

[0052] Step 2: Prediction of small-sample, large-time-scale trajectory information based on recursive convolutional conditional neural processes.

[0053] The present invention proposes Figure 3 The recursive convolutional neural network architecture shown in Figure 1 is composed of a CNP network with a convolutional layer encoder that predicts the trajectory point at the next moment based on the historical trajectory and implements future multi-step prediction through the recursive neural network framework. t Indicates the timestamp at time t, Pos t ,Ang t They represent the position and attitude angle at time t predicted by the network, and the prior data are the timestamps and posture information at time t-1 to tM. t After being jointly input into the CNP network, the mean μ and variance σ of the pose distribution at time t are predicted 2 .

[0054] The Conditional Neural Process (CNP) method combines Bayesian Gaussian processes and deep neural networks, using prior knowledge and training with the help of a neural network framework to achieve good prediction results while using a small dataset and a small model. The network consists of three parts: encoder, aggregator, and decoder:

[0055] The role of the encoder network is to map the conditional input (historical data) to the latent space representation. In order to improve the extraction effect of time series features, the historical data is first passed through a one-dimensional convolution layer, and then the result is input into the multi-layer perceptron (MLP). If the historical data is represented by D hist ={S t-1 ,…,S t-M ,Pos t-1 ,…,Pos t-M ,Ang t-1 ,…,Ang t-M}, the forward transmission process of the encoder can be expressed as:

[0056] Z=h θ (CONV(D hist ))=h θ (D hist *W)

[0057] Where * represents the convolution operation, W is the one-dimensional convolution kernel, CONV(·) represents the convolution layer, h θ (·) represents a multilayer perceptron, θ represents the network weight, and Z is a latent space representation generated by the encoder based on the data.

[0058] The aggregator performs an aggregation operation (such as weighted summation) on the encoder output Z to generate a unified feature representation C:

[0059]

[0060] The decoder is based on the feature C and the query set Generate prediction output, where the query set is the set of timestamps corresponding to the unknown sequence, then the forward propagation process of the decoder can be expressed as:

[0061]

[0062] Among them, Y pred To predict the Gaussian distribution of the pose, it is expressed in terms of mean and variance. The loss function of the CNP network can be transformed into an optimization problem, expressed in terms of negative log-likelihood (NLL):

[0063]

[0064] Among them, p(·|·) is the conditional probability, which represents the CNP according to the historical data Dhist and queryset The distribution of predicted and real data Y true The likelihood.

[0065] Finally, the single-step convolutional CNP network is substituted into the recurrent neural network architecture, and the prediction output of the previous step is added to the conditional input of the current step. The trajectory is predicted in an iterative manner, further improving the accuracy and stability of the prediction.

[0066] Figure 5 The loss function curve of the recursive convolutional CNP network is shown. As the number of training rounds increases, the model gradually converges and the prediction accuracy gradually improves. Figure 6 The prediction effect of the recursive convolutional neural network (CNP) is shown. The blue curve represents the conditional input consisting of 20 sets of historical poses, the red curve and the shaded area represent the Gaussian distribution of the predicted trajectory, and the green curve represents the actual pose data. The actual data are all located near the predicted data and included in the uncertainty area, indicating that the network has good prediction accuracy.

[0067] Step 3: Efficient hierarchical coverage path planning algorithm based on particle swarm optimization.

[0068] Figure 7 The structural block diagram of the path planning algorithm proposed in this invention is shown. First, based on the trajectory prediction generated by the recursive convolutional neural network, an artificial potential field is constructed based on Euclidean distance, probability distribution and time series information. Then, the optimal scanning path is generated through work unit decomposition and coverage path planning. Finally, the local potential field is dynamically updated based on visual recognition information.

[0069] The artificial potential field calculation is divided into the gravitational field U formed by the moving target position att , the repulsive field U formed by the movement disorder rep and the time decay W caused by the temporal characteristics of the predicted trajectory t The purpose of the gravitational field is to attract the laser to scan the target position. First, a piecewise function is used to represent the gravitational force U generated by the Euclidean distance from the workspace to the i-th prediction point. dist ,when Reduce the potential energy when moving away from the target position to avoid the problem of excessive gravitational force when moving away from the target position:

[0070]

[0071] Then, since the network output is a Gaussian distribution of trajectory points, the probability density is introduced as a weight based on the Euclidean distance potential field to introduce the impact of prediction uncertainty, which is specifically expressed as:

[0072]

[0073] Among them, μi,goal and σ i,goal They represent the mean and variance of the Gaussian distribution of target point i, and ζ is the weight coefficient.

[0074] Next, the potential fields calculated based on the n prediction points in the prediction trajectory are superimposed, and the time decay weight is introduced to ensure that the potential field of the prediction point closer to the current moment has a stronger influence. The formula is expressed as:

[0075]

[0076] Here, δ represents the rate at which the weight of the potential field calculated at each prediction point decays over time.

[0077] like Figure 8 The figure shows a reflective prism mounted on a target platform. It is a type of corner cube prism and consists of four faces, three of which are perpendicular to each other at right angles, and the fourth face is the base face with an angle equal to that of the other three faces. Regardless of the point on the base face from which the light is incident, the light will eventually be emitted from the base face, and the outgoing light is parallel to the incident light and in the opposite direction. It is widely used for laser beam ranging. However, due to the limitations of its physical shape and installation position, when the laser is incident on the reflective prism at an angle greater than 45°, it cannot be correctly reflected, resulting in a reduction in the effective scanning angle range of the laser. If scanning is still performed according to the ideal workspace, it will result in ineffective energy loss. Therefore, it is necessary to treat the failure area as an obstacle and introduce a repulsive field:

[0078]

[0079] Where D(q) is the distance to the nearest obstacle, η is the repulsive force increment, Q * Is the range threshold of the obstacle. Within this threshold, the obstacle will produce repulsion, otherwise it will have no effect. The superposition of the gravitational field and the repulsive field forms the final artificial potential field U(m) = U att (m)+U rep (m).

[0080] like Figure 9Figure 1 shows an artificial potential field map generated based on trajectory prediction. Green markers represent trajectory points corresponding to past time steps, and red markers represent target positions predicted by the recursive convolutional neural network (CNP) in the future. The corresponding time sequence is the order of the numbers marked in the figure (for example, number I represents the predicted position at the next moment, and number IV represents the predicted position four time steps later). The dark red shaded area represents the invalid laser scanning area at the current moment (the laser incident angle on the reflective prism in this area is greater than 45° and cannot be properly reflected). From the potential field map, we can see that the predicted point I closest to the current moment has the lowest potential energy, indicating that this is the area where the target reflective prism is most likely to appear. Prediction points farther away from the current moment, such as prediction point V, have higher potential energy, indicating that the probability of the target appearing near these areas is lower. Because it includes time weights, the potential field map needs to be dynamically updated over time. However, due to the long prediction cycle of the neural network, when the predicted trajectory has not yet been updated, it is necessary to optimize the local potential field based on the existing potential field and the temporal and visual information. At the same time, to compensate for the deviation between the predicted and true trajectories, the Gaussian Process Regression (GPR) algorithm is used. Based on the deviation between the historical predicted trajectories and the true trajectories obtained by the visual module, it predicts and compensates for the neural network's future prediction deviations, thereby improving prediction accuracy. Due to its non-parametric structure, Gaussian Process Regression has a fast prediction time and can be used to compensate for the prediction error of the network model in real time. Furthermore, its prediction results are also Gaussian distributed. The prediction process can be expressed as:

[0081]

[0082] Among them, K represents the covariance matrix between different data sets, E train represents the training set of prediction deviation, E test represents the estimated value of the forecast deviation, and the corrected forecast trajectory can be expressed as:

[0083]

[0084] like Figure 10 As shown in (a), the result of segmenting the working unit based on the artificial potential field is shown. The green marked points in the figure represent the target position points recorded in the historical data, and the red marked points represent the future target position points predicted by the recursive convolutional CNP network. The corresponding time sequence is the serial number sequence marked in the figure. The blue covered area represents the working unit divided according to the artificial potential field, and the gray represents the invalid laser scanning area. In order to improve the scanning efficiency, the scanning path planning is only performed in the area where the potential field is lower than a certain threshold. At the same time, the working units of the path planning are divided according to the potential field around the trajectory point to avoid path duplication. The boundaries of the working units are divided based on the pre-set potential field threshold, but the boundary lines are constructed using B-spline lines to ensure smoothness. At the same time, in order to improve the efficiency of scanning, such as those far away from the current moment, Figure 10 To improve the efficiency of local path optimization for the work units associated with predicted points III and IV in (a), it is necessary to increase the boundary threshold of these work units to ensure that the path pre-planning for high-potential work units covers a certain range. In practical applications, the number of work units to be divided for future trajectory points should be adjusted based on hardware computing efficiency and potential field distribution calculation results.

[0085] To achieve efficient and reliable path planning, coverage path planning within each work unit is divided into three steps: path pre-planning, path optimization, and dynamic replanning. The specific measures are described as follows:

[0086] The results of path pre-planning are as follows Figure 10 As shown in (b), the B-spline spirals with equal spacing and full coverage are pre-planned according to the scope of the work unit. The purpose of pre-planning is to Figure 10 (a) generates a uniform and fully covered B-spline spiral. During scanning, as long as the laser falls on the effective reflection area of ​​the target reflective prism, it will be reflected back to the small fast reflector. Therefore, the coverage area of ​​the workspace at each path point is the effective reflection area of ​​the reflective prism at that time. The coverage radius r of the path point can be approximately estimated based on the distance between the target reflective prism and the launch platform. scan .

[0087] The control points of the B-spline spiral are generated in the form of polar coordinates. The geometric center of each working unit is the center of the circle, the positive direction of the Y axis is 0 radians, and a control point is set every ω radians. The shape of each spiral is given by The polar coordinates of the control points are expressed as: r(θ)=nΔr(θ),Δr(θ≤r scan

[0088] Among them, θ represents the angle between the line connecting the control point and the geometric center of the work unit and the positive direction of the Y axis, Δr(θ) represents the distance difference of all control points along the radial direction of the θ angle, n represents the number of turns of the spiral, and the maximum radial distance from the geometric center of the workspace to the boundary and the coverage radius r of the path point are the same. scan The total number of control points of the B-spline curve in each working unit is determined by the ratio of Path optimization based on particle swarm Figure 10 As shown in (c): Although the spiral path generated in the previous step can achieve uniform coverage of the working unit, due to the constant change and uncertainty of the target position, appropriate repeated scanning in the area with high probability of occurrence can improve the probability of alignment. Because the artificial potential field can fully represent the probability distribution of the target's location, the coverage area is weighted based on the strength of the artificial potential field. The density of the scan in the high probability area can be increased:

[0089]

[0090] Among them, A map is the actual physical area of ​​the working unit, U(r,θ) is the strength of the artificial potential field, constants δ1 and δ2 determine the influence of the potential field, A total is the weighted work unit area.

[0091] In addition to considering the coverage of the weighted area, the optimization algorithm also needs to consider the total path length and the actual overlapping area. The cost function is as follows:

[0092]

[0093] Among them, Γ1, Γ2, Γ3 are the weights of the three cost items, A cover The area covered by all scan points in the path.

[0094] The hierarchical particle swarm optimization algorithm is used to optimize the control point position of the B-spline spiral corresponding to each working unit. The outer particle swarm optimization spiral number n loop and the number of new / reduced control points n newp , the number of particles is N outer The inner particle swarm optimizes the positions of all control points P = {(r1,θ1), (r2,θ2),…, (r n ,θ n )}, the number of particles is N inner , the frequency difference of the inner and outer optimization settings is used to improve the optimization efficiency. That is, the inner particle swarm algorithm iterates M times and the outer particle swarm algorithm iterates once. Before starting the optimization, the optimization starting point of each particle position vector is randomly set near the pre-planned path parameters in the form of normal distribution. The position update formula of each particle swarm is:

[0095]

[0096] in, represents the position vector of particle i at time t, Represents the velocity vector of particle i at time t+1. In each round of iteration, the velocity update formula of the particle swarm is:

[0097]

[0098] Among them, w is the inertia weight, c1, c2 are acceleration constants, which represent the dependence of the particle on its own experience and the group experience respectively, r1, r2 are random numbers between [0, 1], which are used to increase the randomness of the search. is the optimal position of particle i, g t is the global optimal position. After multiple iterations, the particle swarm can find the optimal solution in the search space. Figure 10As shown in (c), the optimized path coverage density shows a trend of decreasing radially from the predicted coordinate point to the boundary of the working unit, which ensures that the area with a higher probability of the target reflective prism to obtain more scanning times.

[0099] Dynamic Replanning: When the time corresponding to the next predicted trajectory point is reached, in addition to modifying the time decay weight of the potential field corresponding to each trajectory point, the potential field is fine-tuned based on the actual target position and invalid scanning area calculated by the vision module. After the potential field is updated, the path is replanned based on the new potential field. If the predicted trajectory point position and the surrounding potential field have changed significantly, pre-planning is performed first, followed by particle swarm optimization using the pre-planned B-spline control point as the starting point. Otherwise, particle swarm optimization is performed using the B-spline control point at the previous time step as the starting point.

[0100] The method of the present invention has been verified through simulation experiments and practical applications to have the technical effects claimed by the present invention. The technical problems raised by the present invention have been solved, and the method of the present invention has been verified through practical applications to have the technical effects and practicality claimed by the present invention. The traditional laser alignment method relies on the global cyclic scanning method, and the time consumption of the alignment process is quite random. The present invention uses a binocular camera to accurately identify the target reflection prism, and through active guidance and path planning, drives the laser to quickly approach the area with high target appearance rate, effectively improving the efficiency of laser alignment; and after the traditional method completes one alignment, it needs to re-align again and perform a global scan. The present invention predicts the future position and trajectory of the target reflection prism based on historical data information, and plans the scanning path in advance, which greatly shortens the time interval between the two alignments.

[0101] The method (algorithm) proposed in the present invention for rapid ranging of a space moving target using an optical comb based on a layered guidance strategy is the underlying technical core of the present invention, and various products can be derived based on the algorithm.

[0102] Based on the algorithm (method) proposed in the present invention, a software system for a method of rapid ranging of a space moving target with an optical comb based on a hierarchical guidance strategy is developed using a programming language. The system has program modules corresponding to the steps of the above-mentioned technical solution, and executes the steps of the above-mentioned method of rapid ranging of a space moving target with an optical comb based on a hierarchical guidance strategy during operation.

[0103] The developed system (software) computer program is stored on a computer-readable storage medium. When called by a processor, the computer program is configured to implement the steps of the aforementioned method for rapid ranging of a spatially moving target using an optical comb based on a layered guidance strategy. This materializes the present invention on a carrier, becoming a computer program product.

[0104] A device for rapid ranging of a moving target using an optical comb based on a hierarchical guidance strategy, the device comprising at least one processor and a memory communicatively connected to the at least one processor, wherein the memory stores instructions executable by the at least one processor, and the instructions are executed by the at least one processor to enable the at least one processor to execute the aforementioned method for rapid ranging of a moving target using an optical comb based on a hierarchical guidance strategy, thereby achieving rapid ranging of the moving target.

[0105] Various implementations of the systems and techniques described herein can be realized in digital electronic circuit systems, integrated circuit systems, dedicated ASICs (application specific integrated circuits), computer hardware, firmware, software, and / or combinations thereof. These various implementations can include being implemented in one or more computer programs that can be executed and / or interpreted on a programmable system that includes at least one programmable processor, which can be a special purpose or general purpose programmable processor that can receive data and instructions from a storage system, at least one input device, and at least one output device, and transmit data and instructions to the storage system, the at least one input device, and the at least one output device.

[0106] The computer programs (also referred to as programs, software, software applications, or code) of the present invention include machine instructions for a programmable processor and can be implemented using high-level procedural and / or object-oriented programming languages, and / or assembly / machine languages. As used herein, the terms "machine-readable medium" and "computer-readable medium" refer to any computer program product, device, and / or means (e.g., a magnetic disk, optical disk, memory, programmable logic device (PLD)) for providing machine instructions and / or data to a programmable processor, including machine-readable media that receive machine instructions as machine-readable signals. The term "machine-readable signal" refers to any signal for providing machine instructions and / or data to a programmable processor.

[0107] The above description of the disclosed embodiments will enable one skilled in the art to implement or use the present invention. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the present invention. Therefore, the present invention is not limited to the embodiments shown herein but is to be construed in the widest possible manner consistent with the principles and novel features disclosed herein.

Claims

1. A system for rapid ranging of space moving targets using optical combs based on a layered guidance strategy, characterized in that: The system includes a laser emission and scanning part, an image receiving and processing part, and a processor part; The laser emission and scanning part includes a laser emitter, a large quick reflex mirror, and a large quick reflex mirror driving circuit; the image receiving and processing part includes a small quick reflex mirror, a binocular camera, an image processing module, and a small quick reflex mirror driving circuit; The laser transmitter is used to emit the laser beam used for optical comb ranging. The deflection of the large fast-reflection mirror is used to change the direction of the laser light path, so that the laser scans the target platform and finally aligns with the reflective prism installed on it. The large fast-reflection mirror drive circuit is used for low-level control, driving the fast-reflection mirror to accurately track the reference path given by the main processor. The image of the target reflection prism is reflected to the binocular camera through a small quick reflex mirror, and the imaging result of the binocular camera is input into the image processing module; the small quick reflex mirror is used to reflect external light to the binocular camera. The small quick reflex mirror does not perform large-angle deflection and is only used for fine-tuning and calibration of the reflected light path; the image processing module is used to de-jitter, extract features and fuse the images obtained by the binocular camera, and calculate the relative position with the target. The main processor performs target motion estimation and scanning path planning based on the relative posture information, and inputs the reference path information into the fast mirror drive circuit to implement tracking control.

2. The system for rapid ranging of moving targets using optical combs based on a layered guidance strategy according to claim 1, characterized in that: The binocular camera is a D435i binocular camera; The main processor uses the NVIDIA Jetson Xavier NX high-performance edge computing module to receive relative pose information sent by the image processing module and execute trajectory prediction and path planning algorithm programs.

3. The system for rapid ranging of a moving target using an optical comb based on a layered guidance strategy according to claim 1 or 2, characterized in that: The small fast mirror reflects the external light to the binocular camera and the optical comb ranging module through the beam splitter prism.

4. The system for rapid ranging of moving targets using optical combs based on a layered guidance strategy according to claim 3 is characterized in that: The image receiving and processing part also includes a lens group arranged between the dichroic prism and the binocular camera. The small quick-reflection mirror reflects the external light through the dichroic prism and the laser enters the binocular camera through the lens group. The lens group is used to improve optical quality, control exposure and reduce glare.

5. A method for rapid ranging of a moving target using an optical comb based on a layered guidance strategy, the method comprising: Step 1: Accurate target perception and positioning in complex spatial environments based on multi-sensor fusion perception algorithms An image de-jitter algorithm is introduced to ensure the quality of images of moving targets in space. A Kalman filter and vision-IMU fusion technology are used, and covariance matrix analysis is used to optimize the pose output to suppress drift errors. A D435i camera is used to simultaneously capture RGB and infrared images of the target. The SIFT algorithm is used to efficiently extract feature points from both images, and feature matching is used to improve the recognition accuracy and robustness of image content. Based on the debounced image and matched feature points, the PnP algorithm is used to accurately calculate the relative position of the camera and the target reflective prism. By combining the coordinate transformation matrix between the large quick-reflection mirror of the laser emission module and the camera, the relative position information between the quick-reflection mirror and the reflective prism can be calculated. Based on the calculated three-dimensional relative position, the mechanical scanning angle required to drive the large quick-reflection mirror along its XY axis is calculated, providing the data basis for the quick-reflection mirror's deflection path planning. Step 2: Prediction of small sample and large time scale trajectory information based on recursive convolutional conditional neural process First, a convolutional neural network (CNN) is used as an encoder, taking the historical pose sequence near the current moment as input. The encoder then extracts local temporal information through a one-dimensional convolutional network and maps it into a latent vector h. Subsequently, a multi-layer perceptron (MLP) is used as a decoder, taking the latent vector h and the timestamp of the query pose as input. The decoder ultimately outputs a Gaussian distribution of the relative pose of the reflecting prism and large fast mirror at the future moment. A recurrent neural network (RNN) architecture is introduced into the outer layer of the CNP network, allowing the CNP to recursively predict the pose information of multiple steps in the future. At the same time, the negative log-likelihood (NLL) is used as the loss function to ensure that the predicted distribution can maximize the likelihood of the real data to improve the accuracy of the prediction. Then, based on the predicted position and distribution information of the trajectory points, the potential field distribution map in the workspace is calculated by performing a time-decayed weighted summation of the potential field generated for each predicted point and adding the repulsive field relative to the obstacle. When calculating the potential field of each predicted point, it is necessary to consider the Euclidean distance in the potential field space and the probability density distribution of the predicted point itself. At the same time, by setting a time exponential decay weight for each potential field, predicted points closer to the current time are given a greater weight, thus generating a time-dependent potential field map. Considering that the target reflective prism has a dead zone of incident angle, there may be areas in the workspace that cannot be effectively scanned. A higher potential field value is set for these areas to avoid energy loss caused by ineffective scanning. Step 3: Dynamic planning of B-spline coverage path based on particle swarm optimization algorithm The artificial potential field map obtained in step 2 is used to reflect the probability of the target reflective prism appearing in the workspace. Only areas where the potential field is below a certain threshold are scanned: First, the scanning area is decomposed into multiple simple and non-overlapping work unit areas centered on different prediction points. These areas constitute the workspace for coverage path planning. The work units are divided according to the potential field threshold equipotential lines near each prediction point, and different work units are divided based on different equipotential lines. The construction of the coverage path is achieved by spirals drawn by B-spline curves. The planning process is divided into two steps: pre-planning and online optimization. Pre-planning is used to construct equidistant irregular spirals covering the workspace. The control points of the B-spline are expressed in polar coordinates. The length coordinate spacing of the control points with the same arc is adjusted to generate spirals that meet the requirements. In online optimization, the main indicators of the optimal coverage path are spiral density, coverage area, and path length. The density of the spiral is used to ensure more detailed scanning of areas with a higher probability of target occurrence. The coverage area of ​​the spiral should cover the work unit as much as possible. The starting point of online optimization is the spiral control points generated by pre-planning. By using the particle swarm optimization algorithm, the polar coordinates of each control point and the number of spiral turns are fine-tuned. A loss function is constructed based on the spiral density, path length, and coverage range. Finally, the optimal coverage path within each work unit is optimized. When the prediction results have not been updated, the original potential field needs to be updated according to the target pose updated by the vision module, and the working unit and planned path need to be readjusted; If the original predicted path is relatively accurate, the new path planning can only fine-tune the old B-spline control points; finally, the optimized best path is used as the tracking target and sent to the fast reflection mirror drive module for scanning operation.

6. The method for rapid ranging of a moving target using an optical comb based on a layered guidance strategy according to claim 5, wherein: The precise target perception and positioning in a complex spatial environment based on the multi-sensor fusion perception algorithm described in step 1 is specifically implemented as follows: First, the IMU and camera information are fused to compensate for image jitter caused by the vibration of the on-orbit platform. Then, the infrared camera and RGB camera are fused to adapt to spatial lighting changes. Finally, the relative pose is calculated based on the coordinates of the feature points. The IMU mounted on the camera platform can measure the acceleration and angular velocity of the platform. The motion information of the image is described by the x-direction translation pixel dx, the y-direction translation pixel dy, and the rotation angle da. The high-frequency components of the two motion signals are fused through the Kalman filter to obtain the vibration estimate of the platform. Let the fused state estimate be T, then the fusion process can be expressed as: T k|k =T k|k-1 +K k (z k -H·T k|k-1 ) where x k|k-1 is the predicted state, K k is the Kalman gain, z k is the observation value, and H is the observation matrix. Subsequently, the image pixels captured by the camera can be compensated based on the vibration estimate to eliminate the image distortion caused by the platform vibration: T compensated =T-T motion Using the image information collected by the D435i binocular camera, the SIFT algorithm is used to extract features and obtain the feature point descriptors in the RGB and infrared images, which are expressed as follows: D RGB =SIFT(Image RGB ),D IR =SIFT(Image IR ) Then, by performing feature-level fusion on the feature points of the two, the joint feature vector set D is obtained. fusion =[D RGB ,D IR This feature set integrates all feature information collected by the RGB camera and the infrared camera to improve the accuracy and stability of target recognition under conditions of spatial illumination changes; The PnP algorithm estimates the camera's pose through several known three-dimensional coordinate points and their two-dimensional projections on the image: First, obtain the three-dimensional coordinates P of the preset calibration point on the target platform i And obtain the two-dimensional coordinate p through the camera i , and then solve the camera's external parameters, that is, the rotation matrix R, through direct linear transformation, EPnP and Lvenberg-Marquardt methods. c and T c , three-dimensional coordinates P i To the two-dimensional coordinate p i The projection relationship is: p i =K·[R c |T c ]·P i in, K is the known intrinsic parameter matrix of the camera, which mainly contains the focal length and principal point information of the camera; The small sample and large time scale trajectory information prediction based on the recursive convolution conditional neural process described in step 2 is specifically implemented as follows: A CNP network with a convolutional layer encoder is proposed to predict the trajectory point of the next moment based on the historical trajectory, and a recurrent neural network framework is used to achieve future multi-step prediction; among them, S t Indicates the timestamp at time t, Pos t ,Ang t They represent the position and attitude angle at time t predicted by the network, and the prior data are the timestamps and posture information at time t-1 to tM. t After being jointly input into the CNP network, the mean μ and variance σ of the pose distribution at time t are predicted 2 ; A CNP network with a convolutional layer encoder consists of three parts: encoder, aggregator and decoder: The role of the encoder is to map the conditional input (historical data) to the latent space representation. The historical data is first passed through a one-dimensional convolution layer, and then the result is input into the multi-layer perceptron (MLP). If the historical data is represented by D hist ={S t-1 ,…,S t-M ,Pos t-1 ,…,Pos t-M ,Ang t-1 ,…,Ang t-M }, the forward pass process of the encoder is expressed as: Z=h θ (CONV(D hist ))=h θ (D hist *W) Where * represents the convolution operation, W is the one-dimensional convolution kernel, CONV(·) represents the convolution layer, h θ (·) represents a multilayer perceptron, θ represents the network weight, and Z is a latent space representation generated by the encoder based on the data; The aggregator performs an aggregation operation (such as weighted summation) on the encoder output Z to generate a unified feature representation C: The decoder is based on the feature C and the query set Generate prediction output, the query set is the set of timestamps corresponding to the unknown sequence, then the forward propagation process of the decoder can be expressed as: Among them, Y pred To predict the Gaussian distribution of the pose, it is expressed as mean and variance; the loss function of the CNP network is transformed into an optimization problem, which is expressed as negative log likelihood (NLL): Among them, p(·|·) is the conditional probability, which represents the CNP according to the historical data D hist and queryset The distribution of predicted and real data Y true likelihood; Finally, the single-step convolutional CNP network is substituted into the recurrent neural network architecture, and the prediction output of the previous step is added to the conditional input of the current step to predict the trajectory in an iterative manner; The B-spline coverage path dynamic programming based on the particle swarm optimization algorithm described in step 3 is specifically implemented as follows: First, based on the trajectory prediction generated by the recursive convolutional neural network (CNP), an artificial potential field is constructed based on Euclidean distance, probability distribution, and time series information. Then, the optimal scanning path is generated through work unit decomposition and coverage path planning. Finally, the local potential field is dynamically updated based on visual recognition information. The artificial potential field calculation is divided into the gravitational field U formed by the moving target position att , the repulsive field U formed by the movement disorder rep and the time decay W caused by the temporal characteristics of the predicted trajectory t It consists of three parts; the gravitational field is used to attract the laser to scan the target position. First, a piecewise function is used to represent the gravitational force U generated by the Euclidean distance from the workspace to the i-th prediction point. dist ,when Reduce the potential energy when moving away from the target position to avoid the problem of excessive gravitational force when moving away from the target position: Based on the Euclidean distance potential field, probability density is introduced as a weight to introduce the impact of prediction uncertainty, which is specifically expressed as: Among them, μ i,goal and σ i,goal They represent the mean and variance of the Gaussian distribution of target point i, and ζ is the weight coefficient; Next, the potential fields calculated based on the n prediction points in the prediction trajectory are superimposed, and the time decay weight is introduced to ensure that the potential field of the prediction point closer to the current moment has a stronger influence. The formula is expressed as: Here, δ represents the rate at which the weight of the potential field calculated at each prediction point decays over time. A repulsive field is introduced into the artificial potential field to drive the laser scanning path away from specific areas. When the laser enters the reflective prism at an angle greater than 45°, it cannot be correctly reflected along the original optical path. The scanning action at this time is invalid and needs to be avoided in path planning. By treating these specific areas as obstacles, the laser scanning path is driven away from these areas. The repulsive field can be described as: Where D(q) is the distance to the nearest obstacle, η is the repulsive force increment, Q * is the range threshold of the obstacle. Within this threshold, the obstacle will produce repulsion, otherwise it will have no effect. The superposition of the gravitational field and the repulsive field forms the final artificial potential field U(m) = U att (m)+U rep (m). To compensate for the deviation between the predicted and true trajectories, a Gaussian process regression (GPR) algorithm is used. Based on the deviation between the historical predicted trajectories and the true trajectories obtained by the vision module, the algorithm predicts and compensates for the future prediction deviations of the neural network, thereby improving the prediction accuracy. Gaussian process regression is used to compensate for the prediction error of the network model in real time. Its prediction process can be expressed as: Among them, K represents the covariance matrix between different data sets, E train represents the training set of prediction deviation, E test represents the estimated value of the forecast deviation, and the corrected forecast trajectory can be expressed as: To improve scanning efficiency, only areas with potential fields below a certain threshold are scanned. At the same time, the path planning work units are divided according to the potential fields around the trajectory points to avoid path duplication. The boundaries of the work units are determined by a pre-set potential field threshold, and the boundary lines are constructed using B-spline lines to ensure smoothness. At the same time, to improve the efficiency of local optimization paths at future moments, it is necessary to pre-plan paths for trajectory points at farther moments in the future. Therefore, the boundary threshold of the work units needs to be increased to ensure that the path pre-planning for high potential energy areas can cover a certain range. Taking into account the computational efficiency and potential field distribution, only the trajectory points for the next four steps are divided into work units, and the coverage path within each work unit is planned.

7. The method for rapid ranging of a moving target using an optical comb based on a hierarchical guidance strategy according to claim 6 is characterized in that the coverage path planning within each working unit is performed in three steps: path pre-planning, path optimization, and dynamic replanning, as described in detail below: Path pre-planning: Pre-planning is used to achieve uniform and comprehensive coverage of the spiral line in an irregular workspace. During scanning, the area projected by the target reflection prism into the angle space can be regarded as the coverage area of ​​each path point. The coverage radius r of the scanning point can be approximately estimated based on the distance between the target and the transmitter. scan The control points of the B-spline spiral are generated in the form of polar coordinates. One control point is set for each ω radian. The shape of each spiral is given by The polar coordinates of the control points are expressed as: r(θ)=nΔr(θ),Δr(θ)≤r scan in, θ represents the angle, Δr(θ) represents the radial distance difference of the control point corresponding to the θ angle, n represents the number of turns of the spiral, and the maximum radial distance from the center of the workspace to the boundary is r scan The ratio of Particle swarm-based path optimization: Based on the potential field strength, the coverage area is weighted and integrated to improve the scanning density of high-probability areas: Among them, A map is the original area of ​​the working unit, U(r,θ) is the potential field strength, the δ constant determines the influence of the potential field, A total is the work unit area after weighting by probability density; The total path length and actual overlap area in the optimization algorithm are calculated using the following cost function: Among them, Γ1, Γ2, Γ3 are the weights of the three cost items, A cover is the area covered by all scan points in the path; The hierarchical particle swarm optimization algorithm is used to optimize the control point position, and the outer particle swarm optimization spiral number n loop and the number of new control points n newp The inner particle swarm optimizes the position of the control point P = {(r1,θ1), (r2,θ2),…, (r n ,θ n )}, the frequency difference of the inner and outer optimization settings is used to improve the optimization efficiency; the starting points of the optimization vectors of the N particles are randomly set near the pre-planned path parameters in the form of normal distribution, and the position update formula of the particle swarm is: in, represents the position vector of particle i at time t, represents the velocity vector of particle i at time t+1; the velocity update formula of the particle swarm is Among them, w is the inertia weight, c1, c2 are acceleration constants, which control the dependence of particles on their own experience and group experience respectively, r1, r2 are random numbers between [0, 1], which are used to increase the randomness of the search. is the optimal position of particle i, g t is the global optimal position; after multiple iterations, the particle swarm can find the optimal solution in the search space; Dynamic replanning: When the time corresponding to the next trajectory point is reached, in addition to modifying the time attenuation weight corresponding to the potential field of each trajectory point, the potential field must be fine-tuned based on the target position and invalid scanning area position solved by the vision module. After the potential field is updated, the path is replanned based on the new potential field. At this time, if the predicted trajectory point position and the potential field around it change significantly, pre-planning is performed first, and then particle swarm optimization is performed with the pre-planned B-spline control point as the starting point. Otherwise, particle swarm optimization is performed with the B-spline control point at the previous time step as the starting point.

8. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores a computer program, and the computer program is configured to implement the steps of a method for rapid ranging of a spatial moving target with an optical comb based on a layered guidance strategy according to any one of claims 5 to 7 when called by a processor.

Images