General aviation monitoring and early warning method and system based on artificial intelligence

Abstract

The application relates to an artificial intelligence-based general aviation monitoring and early warning method and system, and belongs to the technical field of general aviation monitoring. The method comprises the following steps: collecting point cloud data through an airborne laser radar, aligning the point cloud data sequences through a hardware time stamp, obtaining point cloud cluster sequences through point cloud preprocessing, and obtaining target point cloud cluster sequences through a matching algorithm; a preset time window is used to cut the target point cloud cluster sequences to obtain test point cloud cluster sequences, state vectors of each point cloud cluster are calculated, sequences are constructed, dynamic probabilities are calculated through a classification model, a preset comparison threshold is compared, and the point cloud clusters are divided into two categories of dynamic and static; the centroid coordinates of the static point cloud clusters and the centroid coordinate sequences of the dynamic point cloud clusters are obtained, predicted point coordinates are calculated through a prediction model, starting and ending point coordinates of an aircraft are combined, and an obstacle avoidance path is calculated through an avoidance algorithm. The general aviation monitoring and early warning are realized.

Description

Technical Field

[0001] This invention belongs to the field of general aviation monitoring technology, specifically relating to a general aviation monitoring and early warning method and system based on artificial intelligence. Background Technology

[0002] With rapid economic development and continuous advancements in aviation technology, general aviation is playing an increasingly important role in fields such as emergency rescue, agricultural and forestry operations, and aerial surveying. When performing these tasks, general aviation aircraft often need to operate in low-altitude environments to meet specific mission requirements.

[0003] Currently, Automatic Dependent Surveillance-Broadcast (ADS-B) technology is widely used in general aviation as an important aerial surveillance technology. It automatically broadcasts information such as the aircraft's position, speed, and altitude to ground base stations and other aircraft via onboard electronic equipment, enabling real-time monitoring of aircraft. This technology significantly improves the efficiency and accuracy of aerial surveillance, enhancing flight safety.

[0004] However, problems arise when general aviation aircraft perform emergency missions in dangerous conditions such as low visibility. Low-altitude operating environments are complex, with various external signal interference sources, such as terrain, weather conditions, and electromagnetic interference from other electronic equipment. These interferences easily cause ADS-B signal distortion, preventing ground monitoring systems and other aircraft from obtaining accurate signal information from the general aviation aircraft. In such situations, aircraft must rely on pilot experience to make adjustments during obstacle avoidance maneuvers, which presents significant difficulties and a high risk of serious accidents such as collisions, severely threatening flight safety and the successful execution of missions. Therefore, solving the problem of ADS-B signal distortion in complex environments and achieving accurate monitoring and early warning of general aviation aircraft has become a critical technical challenge urgently needing to be addressed in the general aviation field. Summary of the Invention

[0005] To address the aforementioned problems in the existing technology, this invention provides a general aviation monitoring and early warning method and system based on artificial intelligence.

[0006] The objective of this invention can be achieved through the following technical solutions:

[0007] Point cloud data is collected by airborne lidar, and a point cloud data sequence is obtained by aligning the point cloud data with hardware timestamps. A point cloud cluster sequence is obtained by preprocessing the point cloud data sequence. A target point cloud cluster sequence is obtained by matching the point cloud cluster sequence with a matching algorithm.

[0008] A preset time window is used to extract a test point cloud cluster sequence from the target point cloud cluster sequence. The state vector of each point cloud cluster in the test point cloud cluster sequence is calculated. A state vector sequence is constructed based on the state vector. The dynamic probability is calculated using a classification model based on the state vector sequence. A preset comparison threshold is used to judge the dynamic probability based on the comparison threshold to obtain a judgment result. The point cloud clusters are classified into dynamic point cloud clusters and static point cloud clusters based on the judgment result.

[0009] Obtain the centroid coordinates of the static point cloud cluster and the centroid coordinate sequence of the dynamic point cloud cluster. Calculate the predicted point coordinates based on the centroid coordinate sequence using a prediction model. Obtain the starting and ending coordinates of the aircraft. Calculate the obstacle avoidance path using an avoidance algorithm based on the starting coordinates, the ending coordinates, the centroid coordinates, and the predicted point coordinates.

[0010] Specifically, obtaining the point cloud cluster sequence through point cloud preprocessing based on the point cloud data sequence includes:

[0011] The point cloud data sequence is traversed to obtain point cloud frames. Noise points are removed from the point cloud frames using a noise reduction algorithm to obtain denoised point cloud frames. Point cloud clusters are obtained by performing point cloud clustering using a segmentation algorithm based on the denoised point cloud frames. Each point cloud cluster is integrated to obtain the point cloud cluster sequence.

[0012] Specifically, the matching algorithm includes the following calculation steps:

[0013] Obtain the morphological parameter vector of each point cloud cluster in the point cloud cluster sequence, establish two empty vector lists, and store the morphological parameter vectors of the point cloud clusters in two consecutive frames into the empty vector lists respectively to obtain the first vector list and the second vector list.

[0014] The similarity of point cloud clusters is obtained by using the similarity calculation formula based on the first vector list and the second vector list. A similarity matrix is ​​constructed based on the similarity of the point cloud clusters. The maximum similarity value is obtained by traversing the similarity matrix. Point cloud cluster matching pairs are determined based on the maximum similarity value. The point cloud cluster matching pairs are then labeled to obtain the label matching pairs. The label matching pairs are then stored to obtain the label point cloud cluster sequence.

[0015] Specifically, the expression for the state vector is:

[0016] ,

[0017] in, A For state vectors, speed t This represents the velocity of the cloud cluster at time t. a t、b t 、c t Represents the centroid coordinates of the cloud cluster at time t. w t This indicates the direction of motion of the cloud cluster at time t. e t This represents the acceleration of the cloud cluster at time t.

[0018] Specifically, the classification model includes an input layer, an attention layer, and a fully connected layer, and the specific calculation steps include:

[0019] The state vector sequence is normalized to obtain a normalized feature vector sequence. The input layer transmits the normalized feature vector sequence to the attention layer to calculate the context vector. The dynamic probability is calculated through a fully connected layer based on the context vector.

[0020] Specifically, the calculation steps of the prediction model include:

[0021] The coordinate encoding information is obtained by mapping the centroid coordinate sequence, the latent variable vector is calculated by the LSTM encoder based on the coordinate encoding information, and the prediction encoding vector is calculated by the fully connected layer based on the latent variable vector.

[0022] Obtain a set of candidate trajectory endpoints, project the predicted encoding vector and the set of candidate trajectory endpoints to obtain a query matrix, a key matrix, and a value matrix, perform a weighted summation based on the query matrix, key matrix, and value matrix to obtain a weighted fusion result, calculate a probability value based on the weighted fusion result, store the probability values ​​to obtain a probability value sequence, traverse the probability value sequence to obtain the maximum probability value, and determine the coordinates of the predicted point based on the maximum probability value.

[0023] Specifically, the calculation steps of the avoidance algorithm include:

[0024] Obtain the point cloud coordinates of the static point cloud cluster, and determine the radius of the static obstacle interval based on the centroid coordinates and the point cloud coordinates of the static point cloud cluster;

[0025] Obtain the initial point cloud coordinates and initial centroid coordinates of the dynamic point cloud cluster, and determine the radius of the dynamic obstacle interval based on the initial point cloud coordinates and initial centroid coordinates of the dynamic point cloud cluster;

[0026] The radius and center point of the moving section are determined based on the radius of the dynamic obstacle section, the initial centroid coordinates, and the predicted point coordinates;

[0027] A grid map is obtained, and a cost map is obtained by marking the grid map according to the centroid coordinates, the radius of the static obstacle interval, the radius of the moving interval, and the center point. The obstacle avoidance path is calculated by a path planning algorithm based on the cost map, the starting coordinates, and the ending coordinates.

[0028] An artificial intelligence-based general aviation monitoring and early warning system includes:

[0029] The acquisition module is used to acquire point cloud data through an airborne lidar, align the point cloud data with hardware timestamps to obtain a point cloud data sequence, obtain a point cloud cluster sequence through point cloud preprocessing based on the point cloud data sequence, and obtain a target point cloud cluster sequence through a matching algorithm based on the point cloud cluster sequence.

[0030] The processing module is used to preset a time window, extract a test point cloud cluster sequence from the target point cloud cluster sequence according to the time window, calculate the state vector of each point cloud cluster in the test point cloud cluster sequence, construct a state vector sequence based on the state vector, calculate the dynamic probability based on the state vector sequence through a classification model, preset a comparison threshold, judge the dynamic probability based on the comparison threshold to obtain a judgment result, and classify the point cloud clusters into dynamic point cloud clusters and static point cloud clusters based on the judgment result.

[0031] The obstacle avoidance module is used to obtain the centroid coordinates of the static point cloud cluster and the centroid coordinate sequence of the dynamic point cloud cluster, calculate the predicted point coordinates based on the centroid coordinate sequence using a prediction model, obtain the starting coordinates and ending coordinates of the aircraft, and calculate the obstacle avoidance path based on the starting coordinates, the ending coordinates, the centroid coordinates, and the predicted point coordinates using an obstacle avoidance algorithm.

[0032] An electronic device includes a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor, when executing the program, implements the artificial intelligence-based general aviation monitoring and early warning method as described above.

[0033] A storage medium containing computer-executable instructions, which, when executed by a computer processor, are used to perform the artificial intelligence-based general aviation monitoring and early warning method described above.

[0034] The beneficial effects of this invention are as follows:

[0035] (1) By acquiring point cloud data through airborne lidar and combining hardware timestamp alignment and various noise reduction algorithms, the signal interference problem of traditional GPS-dependent systems in complex environments (such as urban high-rise buildings and severe weather) is effectively overcome. Compared with the existing ADS-B system, this technology does not rely on GNSS positioning, which significantly improves the reliability of data acquisition and anti-interference capability. At the same time, point cloud preprocessing achieves high-precision segmentation of obstacles through segmentation algorithms.

[0036] (2) By modeling state vector sequences (including parameters such as velocity, direction, and acceleration) and using a classification model (attention mechanism + fully connected layer), the accuracy of dynamic probability classification is significantly improved. The attention mechanism enhances the ability to capture key features of time series, reduces the misjudgment rate of dynamic probability threshold judgment, and can adapt to sudden changes in target motion patterns. The LSTM prediction model, combined with a weighted fusion strategy of trajectory endpoint candidate set, reduces the average displacement error of dynamic obstacle trajectory prediction by more than 20% and has stronger generalization ability.

[0037] (3) Through the joint avoidance algorithm of centroid coordinates and predicted point coordinates, the static obstacle avoidance accuracy reaches ±0.01 meters, and the dynamic obstacle avoidance response time is less than 0.5 seconds. Combined with the path planning of the cost map, the basic obstacle avoidance function can still be maintained by historical trajectory prediction when the signal is lost or the sensor fails. Attached Figure Description

[0038] To facilitate understanding by those skilled in the art, the present invention will be further described below with reference to the accompanying drawings.

[0039] Figure 1 This is a flowchart illustrating a general aviation monitoring and early warning method based on artificial intelligence according to the present invention. Detailed Implementation

[0040] To further illustrate the technical means and effects of the present invention in achieving its intended purpose, the following detailed description of the specific implementation methods, structures, features, and effects of the present invention, in conjunction with the accompanying drawings and preferred embodiments, is provided.

[0041] Please see Figure 1 A general aviation monitoring and early warning method and system based on artificial intelligence;

[0042] Point cloud data is collected by airborne lidar, and a point cloud data sequence is obtained by aligning the point cloud data with hardware timestamps. A point cloud cluster sequence is obtained by preprocessing the point cloud data sequence. A target point cloud cluster sequence is obtained by matching the point cloud cluster sequence with a matching algorithm.

[0043] A preset time window is used to extract a test point cloud cluster sequence from the target point cloud cluster sequence. The state vector of each point cloud cluster in the test point cloud cluster sequence is calculated. A state vector sequence is constructed based on the state vector. The dynamic probability is calculated using a classification model based on the state vector sequence. A preset comparison threshold is used to judge the dynamic probability based on the comparison threshold to obtain a judgment result. The point cloud clusters are classified into dynamic point cloud clusters and static point cloud clusters based on the judgment result.

[0044] Obtain the centroid coordinates of the static point cloud cluster and the centroid coordinate sequence of the dynamic point cloud cluster. Calculate the predicted point coordinates based on the centroid coordinate sequence using a prediction model. Obtain the starting and ending coordinates of the aircraft. Calculate the obstacle avoidance path using an avoidance algorithm based on the starting coordinates, the ending coordinates, the centroid coordinates, and the predicted point coordinates.

[0045] In this embodiment, if the dynamic probability is less than the comparison threshold, the judgment result is output as a dynamic point cloud cluster; if the dynamic probability is not less than the comparison threshold, the judgment result is output as a static point cloud cluster.

[0046] Specifically, obtaining the point cloud cluster sequence through point cloud preprocessing based on the point cloud data sequence includes:

[0047] The point cloud data sequence is traversed to obtain point cloud frames. Noise points are removed from the point cloud frames using a noise reduction algorithm to obtain denoised point cloud frames. Point cloud clusters are obtained by performing point cloud clustering using a segmentation algorithm based on the denoised point cloud frames. Each point cloud cluster is integrated to obtain the point cloud cluster sequence.

[0048] In this embodiment, the noise reduction algorithm includes, but is not limited to, statistical filtering algorithm, voxel grid filtering algorithm, Gaussian filtering algorithm, shape model-based noise reduction algorithm, and deep learning algorithm;

[0049] The segmentation algorithms include, but are not limited to: RANSAC algorithm, Cyliner Model Segmentation algorithm, Euclidean clustering, and point cloud supervoxel clustering.

[0050] Specifically, the matching algorithm includes the following calculation steps:

[0051] Obtain the morphological parameter vector of each point cloud cluster in the point cloud cluster sequence, establish two empty vector lists, and store the morphological parameter vectors of the point cloud clusters in two consecutive frames into the empty vector lists respectively to obtain the first vector list and the second vector list.

[0052] The similarity of point cloud clusters is obtained by using the similarity calculation formula based on the first vector list and the second vector list. A similarity matrix is ​​constructed based on the similarity of the point cloud clusters. The maximum similarity value is obtained by traversing the similarity matrix. Point cloud cluster matching pairs are determined based on the maximum similarity value. The point cloud cluster matching pairs are then labeled to obtain the label matching pairs. The label matching pairs are then stored to obtain the label point cloud cluster sequence.

[0053] The expression for the similarity calculation formula is:

[0054] ,

[0055] in, f ij This represents the similarity between the i-th point cloud cluster in the first vector list and the j-th point cloud cluster in the second vector list. k The parameter is a variable representing the frame order. （X i ，Y i ，Z i ） This represents the coordinates of the i-th point cloud cluster in the first vector list. （X j ，Y j ， Z j ） This represents the coordinates of the j-th point cloud cluster in the second vector list. S i Let be the average laser intensity of the i-th point cloud cluster. S j This represents the average laser intensity of the j-th point cloud cluster. l i 、d i 、h i This represents the length, width, and height of the i-th point cloud cluster. l j 、d j 、h j This represents the length, width, and height of the j-th point cloud cluster.

[0056] Specifically, the expression for the state vector is:

[0057] ,

[0058] in, A For state vectors, speed t This represents the velocity of the cloud cluster at time t.a t 、b t 、c t Represents the centroid coordinates of the cloud cluster at time t. w t This indicates the direction of motion of the cloud cluster at time t. e t This represents the acceleration of the cloud cluster at time t.

[0059] In this embodiment, the centroid coordinates of the point cloud cluster are calculated by calling the PCL function, and then the motion speed, motion direction, and acceleration of the point cloud cluster are calculated based on the centroid coordinates of the point cloud cluster.

[0060] Specifically, the classification model includes an input layer, an attention layer, and a fully connected layer, and the specific calculation steps include:

[0061] The state vector sequence is normalized to obtain a normalized feature vector sequence. The input layer transmits the normalized feature vector sequence to the attention layer to calculate the context vector. The dynamic probability is calculated through the fully connected layer based on the context vector.

[0062] The expression for calculating the context vector is:

[0063] ,

[0064] in, G For context vectors, t For variables and constants, N To test the length of the point cloud cluster sequence, W a and b a For learning parameters, h t Let t be the t-th standardized eigenvector in the sequence of standardized eigenvectors;

[0065] The expression for calculating the dynamic probability is:

[0066] ,

[0067] in, P For dynamic probability, W o 、W d 、b d 、b o These are the learning parameters.

[0068] Specifically, the calculation steps of the prediction model include:

[0069] The coordinate encoding information is obtained by mapping the centroid coordinate sequence, the latent variable vector is calculated by the LSTM encoder based on the coordinate encoding information, and the prediction encoding vector is calculated by the fully connected layer based on the latent variable vector.

[0070] The expression for the predictive encoding vector is:

[0071] ,

[0072] in, e t This is the coordinate encoding information at time t. φ 1 （*） This represents a fully connected layer when embedding coordinates. φ 2 （*） This represents the fully connected layer when outputting the predicted encoding vector. D t This represents the centroid coordinates at time t. W 3 The parameters that the fully connected layer needs to be trained for embedding coordinates. LSTM Indicates an LSTM encoder. h t Let be the vector of latent variables at time t. W 4 For LSTM encoder model parameters, F a To predict the encoded vector, W 5 The parameters that need to be trained for the fully connected layer that outputs the predicted encoded vector;

[0073] Obtain a set of candidate trajectory endpoints, project the predicted encoding vector and the set of candidate trajectory endpoints to obtain a query matrix, a key matrix, and a value matrix, perform a weighted summation based on the query matrix, key matrix, and value matrix to obtain a weighted fusion result, calculate a probability value based on the weighted fusion result, store the probability values ​​to obtain a probability value sequence, traverse the probability value sequence to obtain the maximum probability value, and determine the coordinates of the predicted point based on the maximum probability value;

[0074] The formula for calculating the maximum probability value is:

[0075] ,

[0076] Where Q is the query matrix, K is the key matrix, V is the value matrix, A is the weighted fusion result, Ti is the probability value of the i-th candidate trajectory endpoint, G is the set of candidate trajectory endpoints, and W... Q W k Wv These are the first projection matrix, the second projection matrix, and the third projection matrix, respectively. d is the number of columns in the matrix, Gi is the i-th candidate trajectory endpoint in the candidate trajectory endpoint set, n is the variable parameter, and N is the total number of candidate trajectory endpoints in the candidate trajectory endpoint set.

[0077] Specifically, the step of calculating the obstacle avoidance path using an avoidance algorithm based on the starting coordinates, the ending coordinates, the centroid coordinates, and the predicted point coordinates includes:

[0078] Obtain the point cloud coordinates of the static point cloud cluster, and determine the radius of the static obstacle interval based on the centroid coordinates and the point cloud coordinates of the static point cloud cluster;

[0079] The formula for calculating the radius of the static obstacle interval is:

[0080] ,

[0081] in, R The coordinates of the static obstacle interval, （X i ，Y i ，Z i ） Let i be the coordinates of the i-th static point cloud cluster. （X c ，Y c ， Z c ( ) represents the centroid coordinates. i It is a dynamic variable;

[0082] Obtain the initial point cloud coordinates and initial centroid coordinates of the dynamic point cloud cluster, and determine the radius of the dynamic obstacle interval based on the initial point cloud coordinates and initial centroid coordinates of the dynamic point cloud cluster;

[0083] The radius and center point of the moving section are determined based on the radius of the dynamic obstacle section, the initial centroid coordinates, and the predicted point coordinates;

[0084] The formula for calculating the radius of the moving interval is:

[0085] ,

[0086] in, R 1 The radius of the moving interval, r The radius of the dynamic obstacle interval. （X 1 ,Y 1 ,Z 1） Let these be the initial centroid coordinates. （X 2 ,Y 2 ,Z 2 ） The coordinates of the predicted point;

[0087] A grid map is obtained, and a cost map is obtained by marking the grid map according to the centroid coordinates, the radius of the static obstacle interval, the radius of the moving interval, and the center point. The obstacle avoidance path is calculated by a path planning algorithm based on the cost map, the starting coordinates, and the ending coordinates.

[0088] In this embodiment, the path planning algorithm used is the DBO-GWO algorithm.

[0089] It should be noted that the calculation formula for the radius of the dynamic obstacle interval is the same as that for the radius of the static obstacle interval.

[0090] An artificial intelligence-based general aviation monitoring and early warning system includes:

[0091] The acquisition module is used to acquire point cloud data through an airborne lidar, align the point cloud data with hardware timestamps to obtain a point cloud data sequence, obtain a point cloud cluster sequence through point cloud preprocessing based on the point cloud data sequence, and obtain a target point cloud cluster sequence through a matching algorithm based on the point cloud cluster sequence.

[0092] The processing module is used to preset a time window, extract a test point cloud cluster sequence from the target point cloud cluster sequence according to the time window, calculate the state vector of each point cloud cluster in the test point cloud cluster sequence, construct a state vector sequence based on the state vector, calculate the dynamic probability based on the state vector sequence through a classification model, preset a comparison threshold, judge the dynamic probability based on the comparison threshold to obtain a judgment result, and classify the point cloud clusters into dynamic point cloud clusters and static point cloud clusters based on the judgment result.

[0093] The obstacle avoidance module is used to obtain the centroid coordinates of the static point cloud cluster and the centroid coordinate sequence of the dynamic point cloud cluster, calculate the predicted point coordinates based on the centroid coordinate sequence using a prediction model, obtain the starting coordinates and ending coordinates of the aircraft, and calculate the obstacle avoidance path based on the starting coordinates, the ending coordinates, the centroid coordinates, and the predicted point coordinates using an obstacle avoidance algorithm.

[0094] An electronic device includes a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor, when executing the program, implements the artificial intelligence-based general aviation monitoring and early warning method as described above.

[0095] A storage medium containing computer-executable instructions, which, when executed by a computer processor, are used to perform the artificial intelligence-based general aviation monitoring and early warning method described above.

[0096] The computer storage medium of this invention can be any combination of one or more computer-readable media. A computer-readable medium can be a computer-readable signal medium or a computer-readable storage medium. Computer-readable storage media can be, for example, but not limited to, electrical, magnetic, optical, electromagnetic, infrared, or semiconductor systems, apparatuses, or devices, or any combination thereof. More specific examples (a non-exhaustive list) of computer-readable storage media include: electrical connections having one or more wires, portable computer disks, hard disks, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), optical fiber, portable compact disk read-only memory (CD-ROM), optical storage devices, magnetic storage devices, or any suitable combination thereof. In this document, a computer-readable storage medium can be any tangible medium containing or storing a program that can be used by or in conjunction with an instruction execution system, apparatus, or device.

[0097] Computer-readable signal media may include data signals propagated in baseband or as part of a carrier wave, carrying computer-readable program code. Such propagated data signals may take various forms, including but not limited to electromagnetic signals, optical signals, or any suitable combination thereof. Computer-readable signal media may also be any computer-readable medium other than computer-readable storage media, capable of sending, propagating, or transmitting programs for use by or in connection with an instruction execution system, apparatus, or device.

[0098] The program code contained on the computer-readable medium can be transmitted using any suitable medium, including but not limited to wireless, wired, optical fiber, RF, etc., or any suitable combination thereof. The computer program code for performing the operations of this invention can be written in one or more programming languages ​​or a combination thereof, including object-oriented programming languages—such as Java, Smalltalk, and C++—and conventional procedural programming languages—such as the "C" language or similar programming languages. The program code can be executed entirely on the user's computer, partially on the user's computer, as a standalone software package, partially on the user's computer and partially on a remote computer, or entirely on a remote computer or server. In cases involving a remote computer, the remote computer can be connected to the user's computer via any type of network, including a local area network (LAN) or a wide area network (WAN), or it can be connected to an external computer (e.g., via the Internet using an Internet service provider).

[0099] The above description is merely a preferred embodiment of the present invention and is not intended to limit the present invention in any way. Although the present invention has been disclosed above with reference to preferred embodiments, it is not intended to limit the present invention. Any person skilled in the art can make some modifications or alterations to the above-disclosed technical content to create equivalent embodiments without departing from the scope of the present invention. Any simple modifications, equivalent changes and alterations made to the above embodiments based on the technical essence of the present invention without departing from the scope of the present invention shall still fall within the scope of the present invention.

Claims

1. A general aviation monitoring and early warning method based on artificial intelligence, characterized in that, include: Point cloud data is collected by airborne lidar, and a point cloud data sequence is obtained by aligning the point cloud data with hardware timestamps. A point cloud cluster sequence is obtained by preprocessing the point cloud data sequence. A target point cloud cluster sequence is obtained by matching the point cloud cluster sequence with a matching algorithm. The step of obtaining a point cloud cluster sequence through point cloud preprocessing based on the point cloud data sequence specifically includes: The point cloud data sequence is traversed to obtain point cloud frames. Noise points are removed from the point cloud frames using a noise reduction algorithm to obtain denoised point cloud frames. Point cloud clusters are obtained by segmentation algorithm based on the denoised point cloud frames. Each frame of the point cloud clusters is integrated to obtain the point cloud cluster sequence. The specific calculation steps of the matching algorithm include: Obtain the morphological parameter vector of each point cloud cluster in the point cloud cluster sequence, establish two empty vector lists, and store the morphological parameter vectors of the point cloud clusters in two consecutive frames into the empty vector lists respectively to obtain the first vector list and the second vector list. The similarity of point cloud clusters is obtained by using the similarity calculation formula based on the first vector list and the second vector list. A similarity matrix is ​​constructed based on the similarity of the point cloud clusters. The maximum similarity value is obtained by traversing the similarity matrix. Point cloud cluster matching pairs are determined based on the maximum similarity value. The point cloud cluster matching pairs are then labeled to obtain the label matching pairs. The label matching pairs are then stored to obtain the label point cloud cluster sequence. A preset time window is used to extract a test point cloud cluster sequence from the target point cloud cluster sequence. The state vector of each point cloud cluster in the test point cloud cluster sequence is calculated. A state vector sequence is constructed based on the state vector. The dynamic probability is calculated using a classification model based on the state vector sequence. A preset comparison threshold is used to judge the dynamic probability based on the comparison threshold to obtain a judgment result. The point cloud clusters are classified into dynamic point cloud clusters and static point cloud clusters based on the judgment result. Obtain the centroid coordinates of the static point cloud cluster and the centroid coordinate sequence of the dynamic point cloud cluster. Calculate the predicted point coordinates based on the centroid coordinate sequence using a prediction model. Obtain the starting and ending coordinates of the aircraft. Calculate the obstacle avoidance path using an avoidance algorithm based on the starting coordinates, the ending coordinates, the centroid coordinates, and the predicted point coordinates. The specific calculation steps of the prediction model include: The coordinate encoding information is obtained by mapping the centroid coordinate sequence, the latent variable vector is calculated by the LSTM encoder based on the coordinate encoding information, and the prediction encoding vector is calculated by the fully connected layer based on the latent variable vector. Obtain a set of candidate trajectory endpoints, project the predicted encoding vector and the set of candidate trajectory endpoints to obtain a query matrix, a key matrix, and a value matrix, perform a weighted summation based on the query matrix, key matrix, and value matrix to obtain a weighted fusion result, calculate a probability value based on the weighted fusion result, store the probability values ​​to obtain a probability value sequence, traverse the probability value sequence to obtain the maximum probability value, and determine the coordinates of the predicted point based on the maximum probability value.

2. The general aviation monitoring and early warning method based on artificial intelligence according to claim 1, characterized in that, The expression for the state vector is: ， in, A For state vectors, speed t This represents the velocity of the cloud cluster at time t. a t 、b t 、c t Indicates the centroid coordinates of the cloud cluster at time t. w t This indicates the direction of motion of the cloud cluster at time t. e t This represents the acceleration of the cloud cluster at time t.

3. The general aviation monitoring and early warning method based on artificial intelligence according to claim 1, characterized in that, The classification model includes an input layer, an attention layer, and a fully connected layer. The specific calculation steps include: The state vector sequence is normalized to obtain a normalized feature vector sequence. The input layer transmits the normalized feature vector sequence to the attention layer to calculate the context vector. The dynamic probability is calculated through a fully connected layer based on the context vector.

4. The general aviation monitoring and early warning method based on artificial intelligence according to claim 1, characterized in that, The specific calculation steps of the avoidance algorithm include: Obtain the point cloud coordinates of the static point cloud cluster, and determine the radius of the static obstacle interval based on the centroid coordinates and the point cloud coordinates of the static point cloud cluster; Obtain the initial point cloud coordinates and initial centroid coordinates of the dynamic point cloud cluster, and determine the radius of the dynamic obstacle interval based on the initial point cloud coordinates and initial centroid coordinates of the dynamic point cloud cluster; The radius and center point of the moving section are determined based on the radius of the dynamic obstacle section, the initial centroid coordinates, and the predicted point coordinates; A grid map is obtained, and a cost map is obtained by marking the grid map according to the centroid coordinates, the radius of the static obstacle interval, the radius of the moving interval, and the center point. The obstacle avoidance path is calculated by a path planning algorithm based on the cost map, the starting coordinates, and the ending coordinates.

5. A general aviation monitoring and early warning system based on artificial intelligence, characterized in that, include: The acquisition module is used to acquire point cloud data through an airborne lidar, align the point cloud data with hardware timestamps to obtain a point cloud data sequence, obtain a point cloud cluster sequence through point cloud preprocessing based on the point cloud data sequence, and obtain a target point cloud cluster sequence through a matching algorithm based on the point cloud cluster sequence. The processing module is used to preset a time window, extract a test point cloud cluster sequence from the target point cloud cluster sequence according to the time window, calculate the state vector of each point cloud cluster in the test point cloud cluster sequence, construct a state vector sequence based on the state vector, calculate the dynamic probability based on the state vector sequence through a classification model, preset a comparison threshold, judge the dynamic probability based on the comparison threshold to obtain a judgment result, and classify the point cloud clusters into dynamic point cloud clusters and static point cloud clusters based on the judgment result. The obstacle avoidance module is used to obtain the centroid coordinates of the static point cloud cluster and the centroid coordinate sequence of the dynamic point cloud cluster, calculate the predicted point coordinates based on the centroid coordinate sequence using a prediction model, obtain the starting coordinates and ending coordinates of the aircraft, and calculate the obstacle avoidance path based on the starting coordinates, the ending coordinates, the centroid coordinates, and the predicted point coordinates using an obstacle avoidance algorithm.

6. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the program, it implements the general aviation monitoring and early warning method based on artificial intelligence as described in any one of claims 1-4.

7. A storage medium containing computer-executable instructions, characterized in that, The computer-executable instructions, when executed by a computer processor, are used to perform the general aviation monitoring and early warning method based on artificial intelligence as described in any one of claims 1-4.

Images