A real-time trajectory prediction method and system for UAVs based on low-altitude economy

Abstract

The present invention relates to the field of data processing technology, and in particular to a method and system for predicting the real-time trajectory of a UAV based on low-altitude economy. The method comprises the following steps: collecting data of each position point on the inspection path and the UAV at each position point during historical inspections; obtaining each historical position point corresponding to each position point in the inspection path and the degree of similarity of their postures; obtaining the posture complexity of each position point in the inspection path based on the posture similarity; adaptively inspecting the distance between each group of flight data vectors corresponding to each position point in the inspection path based on the posture complexity; clustering the flight data vectors corresponding to each position point in the inspection path based on the distance to obtain several clusters for each position point in the inspection path, and predicting the flight state of the next position point of the UAV based on the clusters. The present invention improves the accuracy of flight state prediction.

Description

Technical Field

[0001] The present invention relates to the field of data processing technology, and in particular to a method and system for predicting the real-time trajectory of an unmanned aerial vehicle (UAV) based on low-altitude economy. Background Art

[0002] With the rapid development of the low-altitude economy, the large-scale application of drone inspections is becoming increasingly popular. During inspection operations, although drone inspections have pre-set fixed route plans, due to environmental factors, the actual flight trajectory is difficult to completely match the planned trajectory. Therefore, during the actual flight of the drone, its trajectory needs to be predicted so that the actual flight trajectory of the drone is closer to the planned trajectory.

[0003] In related technologies, for example, a Chinese patent application document with authorization publication number CN111461292B discloses a real-time drone trajectory prediction method, including: acquiring data; data preprocessing; generating a data set of various variables of the drone trajectory; using a Markov model to predict the drone status; adding a BN layer according to the drone status to establish a drone trajectory prediction model based on an LSTM network; and predicting the longitude, latitude, and altitude of the drone.

[0004] The above-mentioned related technologies classify the status of drones only based on their speed, without considering environmental changes. For example, when a drone is in two states with the same speed but very different environmental conditions, one is flying with the wind and the other is flying against the wind. Although the speeds are the same, their actual motion states (such as power consumption, flight stability, etc.) are significantly different. Therefore, the one-sidedness of this drone state classification method will result in the next position state predicted by the Markov model not being well applied to trajectory prediction, which in turn causes the subsequent trajectory prediction results to deviate from the actual flight trajectory. Summary of the Invention

[0005] In order to solve the technical problem of classifying the status of a drone only according to its speed without considering environmental conditions, the present invention provides a real-time trajectory prediction method and system for a drone based on low-altitude economy.

[0006] In a first aspect, the present invention provides a method for real-time trajectory prediction of a UAV based on low-altitude economy, which adopts the following technical solutions:

[0007] A method for predicting the real-time trajectory of a UAV based on low-altitude economy, comprising the following steps:

[0008] Collect data from each location point on the inspection route and each location point of the drone during each historical inspection process; based on the data from each location point of the drone during each historical inspection process, obtain the posture feature vector of the drone at each location point during each historical inspection process;

[0009] Obtain each historical position point corresponding to each position point in the inspection path, as well as the first adjacent position point and the second adjacent position point of each historical position point; obtain the posture similarity of each historical position point corresponding to each position point in the inspection path based on the posture feature vectors of the historical position point and the first adjacent position point and the second adjacent position point of the historical position point; obtain the posture complexity of each position point in the inspection path based on the posture similarity;

[0010] Obtain each group of flight data vectors corresponding to each position point in each inspection path; based on the posture complexity, obtain the distance between each group of flight data vectors corresponding to each position point in the inspection path; cluster the flight data vectors corresponding to each position point in the inspection path based on the distance between each group of flight data vectors, and obtain several clusters for each position point in the inspection path; based on the several clusters for each position point in the inspection path, predict the flight state of the UAV during the actual inspection process, and then predict the trajectory of the UAV.

[0011] The innovation of the present invention lies in first obtaining the attitude feature vector of the drone at each position point in the historical inspection process, and based on the attitude feature vector, obtaining the attitude complexity of each position point in the inspection path, reflecting the number of types of flight states at each position point, and then considering environmental change factors to obtain each group of flight data vectors corresponding to each position point in each inspection path; based on the attitude complexity, adjusting the distance between each group of flight data vectors corresponding to each position point in the inspection path, so that the distance between the flight data vectors with more flight states as the fulcrum is larger, so as to subsequently obtain clusters of multiple flight states; finally, clustering the flight data vectors based on the distance metric, and obtaining several clusters for each position point in the inspection path, thereby improving the adaptability and robustness of the drone state division at each position point, and subsequently improving the accuracy of predicting the flight state and trajectory based on the clustering clusters.

[0012] Preferably, the step of obtaining the posture feature vector of the drone at each location point during each historical inspection process includes:

[0013] ;

[0014] Where, Represents the attitude feature vector of the UAV at the j-th position during the i-th historical inspection process; Represents the pitch angle of the UAV at the jth position during the i-th historical inspection process; Represents the roll angle of the UAV at the jth position point in the i-th historical inspection process; Represents the yaw angle of the UAV at the j-th position point in the i-th historical inspection process; Represents the grayscale information entropy of the image collected by the drone at the j-th position during the i-th historical inspection process.

[0015] It is convenient to obtain the posture complexity of each position point in the inspection path based on the posture feature vector.

[0016] Preferably, the step of obtaining each historical position point corresponding to each position point in the inspection path and a first adjacent position point and a second adjacent position point of each historical position point includes:

[0017] Obtain the distance between the kth position point in the inspection path and each position point in any historical inspection process, record the position point in the historical inspection process corresponding to the minimum distance value as a historical position point corresponding to the kth position point in the inspection path, and record the preceding and following adjacent position points of the historical position point in its historical inspection process as the first adjacent position point and the second adjacent position point of the historical position point corresponding to the kth position point in the inspection path.

[0018] Preferably, obtaining the posture similarity between each historical position point and each position point in the inspection path includes:

[0019]

[0020] Where, Represents the similarity between the posture of the hth historical position point and the kth position point in the inspection path; Represents the posture feature vector of the hth historical position point corresponding to the kth position point in the inspection path; The posture feature vector representing the first neighboring position point of the hth historical position point corresponding to the kth position point in the inspection path; The posture feature vector representing the second neighboring position point of the hth historical position point corresponding to the kth position point in the inspection path; Represents the normalized cosine similarity function.

[0021] Preferably, obtaining the posture complexity of each position point in the inspection path includes:

[0022] The standard deviation of the posture similarity of all historical position points corresponding to each position point in the inspection path is used as the posture complexity of each position point in the inspection path.

[0023] The complexity of the posture reflects the number of types of flight states at each location point in the inspection path.

[0024] Preferably, the step of obtaining each set of flight data vectors corresponding to each position point in each inspection path includes:

[0025] The longitude, latitude, horizontal speed, vertical speed, x-axis acceleration, y-axis acceleration, z-axis acceleration, pitch angle, roll angle, yaw angle, and altitude of each historical position point corresponding to each position point in the inspection path are taken as each flight data vector corresponding to each position point in the inspection path;

[0026] The flight data vectors corresponding to each position point in the inspection path are combined in pairs to obtain each group of flight data vectors corresponding to each position point in the inspection path.

[0027] Preferably, obtaining the distance between each group of flight data vectors corresponding to each position point in the inspection path includes:

[0028]

[0029] Where, Represents the distance between the oth group of flight data vectors corresponding to the kth position point in the inspection path; Represents the posture complexity of the kth position point in the inspection path; Represents the number of elements in each flight data vector corresponding to each position point in the inspection path; as well as Represents the value of the mth element of the first flight data vector and the second flight data vector in the oth group of flight data vectors corresponding to the kth position point in the inspection path; Represents the information entropy of the mth element in all flight data vectors corresponding to the kth position point in the inspection path.

[0030] This allows more types of flight states to be formed at locations with greater posture complexity.

[0031] Preferably, the step of obtaining a plurality of clusters for each location point in the inspection path includes:

[0032] The neighborhood radius ε is preset, and the minimum number of points minpts within the radius is set. Based on the distance between each group of historical location points corresponding to each location point in the inspection path, DBSCAN clustering is used to cluster the flight data vector corresponding to each location point in the inspection path, and several clusters are obtained for each location point in the inspection path.

[0033] Preferably, the method of predicting the flight state of the UAV during the actual inspection process based on a plurality of clusters at each location point in the inspection path includes:

[0034] During the actual inspection process of the drone, the flight data vector of the drone's current position is obtained; the distance between the drone's current position and each position point in the inspection path is obtained, and the position point in the inspection path corresponding to the minimum distance is recorded as the target position point;

[0035] According to the flight data vector of the current position of the UAV, the KNN algorithm with K=5 is used in the cluster cluster of the target position to obtain the cluster to which the flight data vector of the current position of the UAV belongs; the cluster to which the flight data vector of the current position of the UAV belongs is input into the Markov model to predict the flight status of the next position of the UAV.

[0036] Environmental change factors are taken into account, which improves the accuracy of flight status prediction.

[0037] In a second aspect, the present invention provides a real-time trajectory prediction system for UAVs based on low-altitude economy, which adopts the following technical solutions:

[0038] A real-time trajectory prediction system for UAVs based on low-altitude economy includes: a processor and a memory, wherein the memory stores computer program instructions. When the computer program instructions are executed by the processor, the real-time trajectory prediction method for UAVs based on low-altitude economy is implemented.

[0039] By adopting the above technical solution, the above-mentioned method for real-time trajectory prediction of drones based on low-altitude economy is generated into a computer program and stored in a memory so as to be loaded and executed by a processor, thereby making a terminal device based on the memory and the processor for easy use.

[0040] The present invention has the following technical effects: the present invention first obtains the posture complexity of each position point in the inspection path based on the difference in historical inspection data at the position point, reflecting the number of types of flight states at each position point; then obtains each group of flight data vectors corresponding to each position point in each inspection path, representing the flight state of the drone, and adjusts the distance between each group of flight data vectors corresponding to each position point in the inspection path based on the posture complexity, so that the distance between the flight data vectors of position points with more flight states is larger, so as to obtain more clusters, thereby improving the adaptability and robustness of the drone state division at each position point, and subsequently predicting the flight state and trajectory during the actual inspection process based on the clusters. BRIEF DESCRIPTION OF THE DRAWINGS

[0041] Figure 1 This is a flow chart of a method for real-time trajectory prediction of a UAV based on low-altitude economy in an embodiment of the present invention. DETAILED DESCRIPTION

[0042] The technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, but not all of the embodiments.

[0043] The embodiment of the present invention discloses a method for predicting the real-time trajectory of a UAV based on low-altitude economy, referring to Figure 1 , including steps S1 to S4:

[0044] S1: Collect data at each location point on the inspection route and collect data at each location point during the historical inspection process.

[0045] In an embodiment of the present invention, each position point on the inspection path is obtained, and the longitude, latitude, horizontal speed, vertical speed, x-axis acceleration, y-axis acceleration, z-axis acceleration, pitch angle, roll angle, yaw angle and altitude of the drone at each position point during each historical inspection are obtained, as well as the image collected at each position.

[0046] S2: Based on the data of the drone at each position point during the historical inspection process, obtain the posture feature vector of the drone at each position point during each historical inspection process; obtain each historical position point corresponding to each position point in the inspection path; obtain the posture similarity of each historical position point corresponding to each position point in the inspection path; based on the posture similarity, obtain the posture complexity of each position point in the inspection path.

[0047] It should be noted that under normal circumstances, drones should maintain stable flight during inspections. When the flight environment (wind speed, airflow or obstacles) changes, it is usually necessary to adjust the drone's power system to maintain the stability of the drone's position, otherwise it will deviate from the planned trajectory. The adjustment of the drone's power system will cause the drone's attitude angle to change to a certain extent. Since the camera carried by the drone is usually fixed on the fuselage and strictly bound to the fuselage attitude, when the fuselage attitude angle changes, the camera will tilt and rotate synchronously with the fuselage, resulting in a change in the shooting angle, which means that the information entropy of the image collected by the drone changes. Therefore, the pitch angle, roll angle and yaw angle of the drone during the inspection process and the information entropy of the collected image are used to construct the attitude feature vector of the drone during the inspection process, which is used to describe the attitude characteristics of the drone during the inspection process.

[0048] In this embodiment of the present invention, the posture feature vector of the drone at each location point during each historical inspection process is obtained:

[0049]

[0050] Where, Represents the attitude feature vector of the UAV at the j-th position during the i-th historical inspection process; Represents the pitch angle of the UAV at the jth position during the i-th historical inspection process; Represents the roll angle of the UAV at the jth position point in the i-th historical inspection process; Represents the yaw angle of the UAV at the j-th position point in the i-th historical inspection process; Represents the grayscale information entropy of the image collected by the drone at the j-th position during the i-th historical inspection process;

[0051] It should be noted that when the inspection path of the drone is well specified and there is no significant change in the environment, it is theoretically believed that the flight state of the drone at the same position on the path during all historical inspections should be similar; however, if the drone is disturbed by the environment during a certain inspection, causing the drone to fly to that position in the inspection path, its flight state deviates from the flight state of the drone when it flew to that position in the historical inspection process, resulting in abnormal fluctuations in the flight state; for any position point in the inspection path, the greater the difference between the posture similarity of the position point when passing through it in all historical inspections and the posture similarity of the previous and next position points, it means that the drone may pass through this position point in the inspection path with different postures, and the posture complexity of this position point in the inspection path is higher.

[0052] In an embodiment of the present invention, the distance between the kth position point in the inspection path and each position point in any historical inspection process is obtained, and the position point in the historical inspection process corresponding to the minimum distance value is recorded as a historical position point corresponding to the kth position point in the inspection path, and the preceding and following adjacent position points of the historical position point in its historical inspection process are recorded as the first adjacent position point and the second adjacent position point of the historical position point corresponding to the kth position point in the inspection path;

[0053] Similarly, each historical position point corresponding to each position point in the inspection path is obtained, and the first adjacent position point and the second adjacent position point of each historical position point corresponding to each position point in the inspection path are obtained.

[0054] Get the posture similarity of each historical position point corresponding to each position point in the inspection path:

[0055]

[0056] Where, Represents the similarity between the posture of the hth historical position point and the kth position point in the inspection path; Represents the posture feature vector of the hth historical position point corresponding to the kth position point in the inspection path; Represents the first neighboring point of the hth historical point corresponding to the kth point in the inspection path Posture characteristics vector; The posture feature vector representing the second neighboring position point of the hth historical position point corresponding to the kth position point in the inspection path; Represents the normalized cosine similarity function. It should be noted that, Equal to the cosine similarity between A and B plus 1 divided by 2; The larger the value of , the more similar the posture of the UAV when it flies to the kth position point in the inspection path during the historical inspection process is to the posture when it flies to the positions before and after the kth position point. The larger the value of , the more similar the UAV is when it passes through the kth position point and the positions before and after the kth position point in the inspection path in the historical inspection process with the same posture.

[0057] It should be noted that if the drone passes through any position point in the inspection path and its previous and next position points with the same posture in all historical inspection processes, then the posture complexity of the position point in the inspection path will be smaller. Conversely, if the drone passes through any position point in the inspection path and its previous and next position points with different postures in all historical inspection processes, then the posture complexity of the position point in the inspection path will be greater.

[0058] The standard deviation of the posture similarity of all historical position points corresponding to each position point in the inspection path is used as the posture complexity of each position point in the inspection path.

[0059] S3: Obtain the flight data vector corresponding to each position point in the inspection path, and obtain the distance between each group of flight data vectors corresponding to each position point in the inspection path based on the posture complexity of each position point in the inspection path and the flight data vector; based on the distance, cluster the flight data vector corresponding to each position point in the inspection path to obtain several clusters for each position point in the inspection path.

[0060] It should be noted that, since the flight status of the drone is different when it flies to different positions in the inspection path, if the DBSCAN clustering algorithm is directly used to cluster the drone status at all positions on the inspection path, the unique flight status of different positions will be mixed up, resulting in the clustering result being unable to accurately reflect the actual flight characteristics of each positioning point, making the subsequent state prediction of the drone inaccurate. Therefore, the present invention uses the DBSCAN clustering algorithm to cluster the flight status of all historical position points corresponding to each position point in the inspection path;

[0061] However, when a fixed distance threshold is used to cluster the flight status of all historical position points corresponding to each position point in the inspection path, if the number of types of flight status at all historical position points corresponding to any position point in the inspection path is greater, then the use of a fixed threshold may cause historical position points originally belonging to different flight statuses to be merged into the same cluster, affecting the subsequent prediction of flight status. Therefore, the present invention dynamically adjusts the distance measurement method between the historical position points corresponding to each position point in the inspection path according to the posture complexity of each position point in the inspection path, so that the actual distance between the historical position points of different flight statuses exceeds the fixed distance threshold, avoiding the merging of clusters that should have been independent, thereby making the clustering result more in line with the actual number of states, that is, the greater the posture complexity of any position point in the inspection path, the more types of flight status there are at all historical position points corresponding to the position point in the inspection path, and at this time, the distance between the historical position points corresponding to each position point in the inspection path is widened.

[0062] In an embodiment of the present invention, the longitude, latitude, horizontal speed, vertical speed, x-axis acceleration, y-axis acceleration, z-axis acceleration, pitch angle, roll angle, yaw angle and altitude of each historical position point corresponding to each position point in the inspection path flown by the drone are used as each flight data vector corresponding to each position point in the inspection path.

[0063] The flight data vectors corresponding to each position point in the inspection path are combined in pairs to obtain each group of flight data vectors corresponding to each position point in the inspection path.

[0064] Get the distance between each set of flight data vectors corresponding to each position point in the inspection path:

[0065]

[0066] Where, Represents the distance between the oth group of flight data vectors corresponding to the kth position point in the inspection path; Represents the posture complexity of the kth position point in the inspection path; Represents the number of elements in each flight data vector corresponding to each position point in the inspection path; Represents the value of the mth element of the first flight data vector in the oth group of flight data vectors corresponding to the kth position point in the inspection path; Represents the value of the mth element of the second flight data vector in the oth group of flight data vectors corresponding to the kth position point in the inspection path; Represents the information entropy of the mth element in all flight data vectors corresponding to the kth position point in the inspection path;

[0067] represents the difference of the mth element in the oth group of flight data vectors corresponding to the kth position point in the inspection path. The larger its value is, the greater the distance between the oth group of flight data vectors is;

[0068] The larger the value of is, the greater the fluctuation of the mth element when the drone flies to the kth position point in the inspection path during all historical inspections. When the value of is larger, the distance between the oth group of flight data vectors is larger and the credibility is higher;

[0069] The larger the value of , the more types of flight states there are when the drone flies to the kth position point in the inspection path during all historical inspections. This means that the actual position points corresponding to the drone flying to the kth position point in the inspection path during the historical inspections are more dispersed in spatial distribution. At this time, the distance between the oth group of flight data vectors corresponding to the kth position point in the inspection path is larger, making it easier for the distance to exceed the distance threshold preset by the DBSCAN clustering algorithm, thereby forming more types of flight states at the kth position point in the inspection path.

[0070] The preset neighborhood radius ε=5, the minimum number of points within the neighborhood radius minpts=15, and based on the distance between each group of historical location points corresponding to each location point in the inspection path, DBSCAN clustering is used to cluster the flight data vector corresponding to each location point in the inspection path, and several clusters are obtained for each location point in the inspection path.

[0071] S4: Obtain the flight data vector of the current position of the drone during the actual inspection process. Based on several clusters of each position in the inspection path and the flight data vector of the current position of the drone, predict the flight state of the next position of the drone, and then predict the trajectory of the drone.

[0072] In an embodiment of the present invention, during the actual inspection process of the UAV, a flight data vector of the current position point of the UAV is obtained;

[0073] Obtain the distance between the drone's current location and each location in the inspection path. Record the location in the inspection path corresponding to the minimum distance as the target location. Based on the flight data vector of the drone's current location, use the KNN algorithm with K=5 in the clustering cluster of the target location to obtain the cluster to which the flight data vector of the drone's current location belongs.

[0074] The flight data vector of the current position of the UAV belongs to the cluster, which is input into the Markov model to predict the flight state of the UAV at the next position. Finally, the trajectory of the UAV is predicted based on the content of the Chinese patent application document with the authorization announcement number CN111461292B.

[0075] The above are all preferred embodiments of the present invention, and are not intended to limit the scope of protection of the present invention. Therefore, any equivalent changes made based on the structure, shape, and principle of the present invention should be included in the scope of protection of the present invention.

Claims

1. A real-time trajectory prediction method for UAVs based on low-altitude economy, characterized in that: include: Collect data on each location point on the inspection route and the drone's position at each location point during each historical inspection; According to the data of the UAV at each location point during each historical inspection process, the posture feature vector of the UAV at each location point during each historical inspection process is obtained; Obtain each historical location point corresponding to each location point in the inspection path and the first adjacent location point and the second adjacent location point of each historical location point; Based on the posture feature vectors of the historical position point and the first and second adjacent position points of the historical position point, obtaining the posture similarity of each historical position point corresponding to each position point in the inspection path; based on the posture similarity, obtaining the posture complexity of each position point in the inspection path; Obtain each set of flight data vectors corresponding to each position point in each inspection path; based on the posture complexity, obtain the distance between each set of flight data vectors corresponding to each position point in the inspection path; Based on the distance between each group of flight data vectors, the flight data vectors corresponding to each position point in the inspection path are clustered to obtain a number of clusters for each position point in the inspection path; based on the several clusters for each position point in the inspection path, the flight state of the UAV during the actual inspection process is predicted, and then the trajectory of the UAV is predicted; The obtaining of the distance between each group of flight data vectors corresponding to each position point in the inspection path includes: Where, Represents the distance between the oth group of flight data vectors corresponding to the kth position point in the inspection path; Represents the posture complexity of the kth position point in the inspection path; Represents the number of elements in each flight data vector corresponding to each position point in the inspection path; as well as Represents the value of the mth element of the first flight data vector and the second flight data vector in the oth group of flight data vectors corresponding to the kth position point in the inspection path; Represents the information entropy of the mth element in all flight data vectors corresponding to the kth position point in the inspection path.

2. The method for real-time trajectory prediction of a UAV based on low-altitude economy according to claim 1, characterized in that: The acquisition of the attitude feature vector of the drone at each location point during each historical inspection process includes: ； Where, Represents the attitude feature vector of the UAV at the j-th position during the i-th historical inspection process; Represents the pitch angle of the UAV at the jth position during the i-th historical inspection process; Represents the roll angle of the UAV at the jth position point in the i-th historical inspection process; Represents the yaw angle of the UAV at the j-th position point in the i-th historical inspection process; Represents the grayscale information entropy of the image collected by the drone at the j-th position during the i-th historical inspection process.

3. The method for real-time trajectory prediction of a UAV based on low-altitude economy according to claim 1, characterized in that: The step of obtaining each historical position point corresponding to each position point in the inspection path and a first adjacent position point and a second adjacent position point of each historical position point includes: Obtain the distance between the kth position point in the inspection path and each position point in any historical inspection process, record the position point in the historical inspection process corresponding to the minimum distance value as a historical position point corresponding to the kth position point in the inspection path, and record the preceding and following adjacent position points of the historical position point in its historical inspection process as the first adjacent position point and the second adjacent position point of the historical position point corresponding to the kth position point in the inspection path.

4. The method for real-time trajectory prediction of a UAV based on low-altitude economy according to claim 1, characterized in that: The step of obtaining the posture similarity of each historical position point corresponding to each position point in the inspection path includes: Where, Represents the similarity between the posture of the hth historical position point and the kth position point in the inspection path; Represents the posture feature vector of the hth historical position point corresponding to the kth position point in the inspection path; The posture feature vector representing the first neighboring position point of the hth historical position point corresponding to the kth position point in the inspection path; The posture feature vector representing the second neighboring position point of the hth historical position point corresponding to the kth position point in the inspection path; Represents the normalized cosine similarity function.

5. The method for real-time trajectory prediction of a UAV based on low-altitude economy according to claim 1, characterized in that: The step of obtaining the posture complexity of each position point in the inspection path includes: The standard deviation of the posture similarity of all historical position points corresponding to each position point in the inspection path is used as the posture complexity of each position point in the inspection path.

6. The method for real-time trajectory prediction of a UAV based on low-altitude economy according to claim 1, characterized in that: The step of obtaining each set of flight data vectors corresponding to each position point in each inspection path includes: The longitude, latitude, horizontal speed, vertical speed, x-axis acceleration, y-axis acceleration, z-axis acceleration, pitch angle, roll angle, yaw angle, and altitude of each historical position point corresponding to each position point in the inspection path are taken as each flight data vector corresponding to each position point in the inspection path; The flight data vectors corresponding to each position point in the inspection path are combined in pairs to obtain each group of flight data vectors corresponding to each position point in the inspection path.

7. The method for real-time trajectory prediction of a UAV based on low-altitude economy according to claim 1, characterized in that: The step of obtaining a plurality of clusters for each location point in the inspection path includes: The neighborhood radius ε is preset, and the minimum number of points minpts within the radius is set. Based on the distance between each group of historical location points corresponding to each location point in the inspection path, DBSCAN clustering is used to cluster the flight data vector corresponding to each location point in the inspection path, and several clusters are obtained for each location point in the inspection path.

8. The method for real-time trajectory prediction of a UAV based on low-altitude economy according to claim 1, characterized in that: The method predicts the flight status of the UAV during the actual inspection process based on the multiple clusters of each location point in the inspection path, including: During the actual inspection process of the drone, the flight data vector of the drone's current position is obtained; the distance between the drone's current position and each position point in the inspection path is obtained, and the position point in the inspection path corresponding to the minimum distance is recorded as the target position point; According to the flight data vector of the current position of the UAV, the KNN algorithm with K=5 is used in the cluster cluster of the target position to obtain the cluster to which the flight data vector of the current position of the UAV belongs; the cluster to which the flight data vector of the current position of the UAV belongs is input into the Markov model to predict the flight status of the next position of the UAV.

9. A real-time trajectory prediction system for UAVs based on low-altitude economy, characterized by: include: A processor and a memory, wherein the memory stores computer program instructions, and when the computer program instructions are executed by the processor, a method for real-time trajectory prediction of a UAV based on low-altitude economy according to any one of claims 1 to 8 is implemented.

Images