An unmanned vehicle target attribute identification method for dynamic targets in a field environment
By receiving situational information in the unmanned vehicle and using the Kalman filter algorithm for trajectory prediction, the problem of insufficient accuracy in dynamic target recognition in the field environment of the unmanned vehicle is solved, and high-precision dynamic target attribute recognition is achieved.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- ZHONGBING INTELLIGENT INNOVATION RES INST CO LTD
- Filing Date
- 2025-04-30
- Publication Date
- 2026-07-14
AI Technical Summary
Autonomous vehicles struggle to effectively identify dynamic targets in wilderness environments, and existing technologies suffer from insufficient accuracy due to time delays.
The command and control terminal collects situational information in real time and sends it to the unmanned vehicle to establish a situational sequence. The Kalman filter algorithm is used for trajectory prediction and matching, and the reconnaissance coordinates at the reconnaissance time are used to identify target attributes.
It improves the accuracy of dynamic target attribute recognition, reduces the impact of time delay on recognition, and optimizes the prediction accuracy of the Kalman filter model by setting preset conditions and situation point thresholds.
Smart Images
Figure CN120451202B_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of unmanned vehicle situational awareness technology, and in particular to a method for unmanned vehicle target attribute recognition of dynamic targets in a field environment. Background Technology
[0002] With the rapid development of autonomous vehicle technology, its application in various fields is becoming increasingly widespread. Traditional target attribute recognition methods for autonomous vehicles mainly involve acquiring target images and then identifying the target attributes through these images. However, in complex outdoor environments, autonomous vehicles often lack the resources to effectively acquire target images, and the ever-changing outdoor environment makes target recognition based on images acquired by autonomous vehicles extremely difficult.
[0003] The above problems can be solved by remotely collecting target information and transmitting it to the autonomous vehicle. Existing technology involves remotely collecting situational information, including target coordinates and attribute information, in real time and transmitting it to the autonomous vehicle. After receiving this situational information, the autonomous vehicle performs data analysis and comparison with the target information it has detected.
[0004] However, existing technologies have significant limitations in practical applications. On the one hand, they are mainly designed for static targets and are difficult to adapt to the dynamic changes of targets in the field environment. On the other hand, time delays are unavoidable in the process of updating and issuing command and control situation information. This leads to a certain error between the target information actually detected by the unmanned vehicle and the received situation target information, which in turn reduces the accuracy of target attribute identification. Summary of the Invention
[0005] Based on the above analysis, the present invention aims to provide a method for unmanned vehicle target attribute recognition of dynamic targets in a field environment, in order to solve the problems that existing unmanned vehicle recognition methods are difficult to adapt to dynamic targets and have insufficient recognition accuracy due to time delays.
[0006] On one hand, embodiments of the present invention provide a method for unmanned vehicle target attribute recognition of dynamic targets in a field environment, the method comprising:
[0007] The command and control terminal collects situational information in real time and sends it to the unmanned vehicle. The situational information includes the coordinates and attribute information of each target.
[0008] The autonomous vehicle receives situational information and forms situational points based on the situational information and the time of reception. It then generates a situational sequence by generating all situational points of the same target within a preset time period.
[0009] After the unmanned vehicle detects any target, it acquires the reconnaissance information of the target, including the reconnaissance coordinates and the reconnaissance time.
[0010] Determine whether the number of situation points in the situation sequence of each target meets the preset conditions. When the preset conditions are met, perform trajectory prediction for each target based on the situation sequence of the target to obtain the predicted coordinates of each target at the reconnaissance time.
[0011] The predicted coordinates of all targets at the reconnaissance time are matched with the reconnaissance coordinates to obtain the matched predicted coordinates. The attribute information of the target corresponding to the matched predicted coordinates is used as the attribute information of the reconnaissance target.
[0012] As a further improvement to this application, the preset condition is: the number of situation points in the target's situation sequence is greater than or equal to the situation point threshold.
[0013] If the preset conditions are not met, the coordinates of the last situation point in each target situation sequence will be used as the predicted coordinates of the target at the reconnaissance time.
[0014] As a further improvement to this application, matching the predicted coordinates of all targets at the reconnaissance time with the reconnaissance coordinates includes:
[0015] Calculate the distance between the predicted coordinates of all targets at the reconnaissance time and the reconnaissance coordinates;
[0016] The target with the shortest distance between the predicted coordinates at the reconnaissance time and the reconnaissance coordinates is selected as the matching target.
[0017] As a further improvement to this application, the attribute information of the reconnaissance target determined based on the matching results includes:
[0018] The attribute information of the matching target is assigned to the reconnaissance target, thereby obtaining the attribute information of the reconnaissance target.
[0019] As a further improvement of this application, if the number of situation points in the target's situation sequence meets a preset threshold, then the Kalman filter algorithm is used to predict the target's trajectory to obtain the predicted coordinates of the target at the reconnaissance time. Specifically, this includes the following steps:
[0020] Input all situation points of the situation sequence of targets greater than or equal to the situation point threshold into the initial Kalman filter model, and update the Kalman gain of the initial Kalman filter model based on the situation sequence to obtain the final Kalman model;
[0021] The situation information of the last situation point in the situation sequence, as well as the time interval between the reception time and the reconnaissance time of the last situation point, are input into the final Kalman model to obtain the target's coordinates at the reconnaissance time.
[0022] As a further improvement to this application, the final Kalman model is obtained by updating the Kalman gain of the initial Kalman filter model based on the situation sequence, including:
[0023] Starting from the second situation point, perform the following operations sequentially for each situation point in the situation sequence:
[0024] Input the coordinates of the previous situation point and the receiving time into the Kalman filter model to calculate the estimated coordinates at the current time.
[0025] The coordinates of the current situation point are used as coordinate observations, and the Kalman gain is updated based on the coordinate observations and the estimated coordinates.
[0026] The final Kalman model is obtained by using the last updated Kalman gain as the final gain of the Kalman filter model.
[0027] As a further improvement of this application, the situation information of the last situation point in the situation sequence and the time interval between the reception time and the reconnaissance time of the last situation point are input into the final Kalman model to obtain the target's coordinates at the reconnaissance time as shown in the following formula.
[0028]
[0029]
[0030] in, , The coordinates of the target at the time of reconnaissance. , The coordinates of the last situation point. The target velocity at the moment of reception of the last situation point. The turning angle at the moment of receiving the last situation point. The time interval between the reception time of the last situation point and the reconnaissance time. For Kalman gain.
[0031] As a further improvement to this application, the target velocity at the last state point reception time is shown in the following formula;
[0032] ；
[0033] ；
[0034] ;
[0035] in, The x-axis component of the target velocity at the last state point reception time. The target velocity y-axis component is the value received at the last state point. The time interval between the last situation point and the previous situation point. , Set the coordinates of the last state point at the previous moment;
[0036] The turning angle at the moment of receiving the last situation point is shown in the following formula;
[0037] .
[0038] As a further improvement to this application, the coordinates of the situation points in the situation sequence are latitude, longitude, and altitude coordinates; before updating the Kalman gain of the initial Kalman filter model based on the situation sequence to obtain the final Kalman model, the method further includes:
[0039] Convert the coordinates of the situation points in the situation sequence to rectangular coordinates.
[0040] As a further improvement to this application, the situation point threshold is 10.
[0041] Compared with the prior art, the present invention can achieve at least one of the following beneficial effects:
[0042] 1. This invention uses an unmanned vehicle to receive situational information sent by the command and control terminal and establish a situational sequence. It uses the Kalman filter algorithm to predict the trajectory of the target and matches it with the reconnaissance coordinates at the reconnaissance time. This can accurately identify the attribute information of dynamic targets, fully consider the dynamic characteristics of the targets, and improve the accuracy of target attribute identification through prediction and matching.
[0043] 2. Based on the situation sequence, this invention effectively solves the problem of mismatch between situation target information and actual reconnaissance target information caused by time delay by introducing preset conditions and situation point thresholds. When the number of situation points does not meet the preset conditions, the coordinates of the last situation point are used as the prediction coordinates, reducing the impact of time delay on target attribute identification. When the number of situation points in the target's situation sequence meets the preset conditions, the Kalman filter algorithm is used for trajectory prediction, and the prediction accuracy of the Kalman filter model is further improved by converting the situation point coordinates in the situation sequence into rectangular coordinates.
[0044] In this invention, the above-described technical solutions can be combined with each other to achieve more preferred combinations. Other features and advantages of this invention will be set forth in the following description, and some advantages may become apparent from the description or be learned by practicing the invention. The objects and other advantages of this invention can be realized and obtained from what is particularly pointed out in the description and drawings. Attached Figure Description
[0045] The accompanying drawings are for illustrative purposes only and are not intended to limit the invention. Throughout the drawings, the same reference numerals denote the same parts.
[0046] Figure 1 is a flowchart illustrating a method for identifying the target attributes of an unmanned vehicle for dynamic targets in a field environment, according to an embodiment of the present invention. Detailed Implementation
[0047] Preferred embodiments of the present invention will now be described in detail with reference to the accompanying drawings, which constitute a part of this application and are used together with the embodiments of the present invention to illustrate the principles of the present invention, but are not intended to limit the scope of the invention.
[0048] A specific embodiment of the present invention discloses a method for identifying the target attributes of dynamic targets by unmanned vehicles in a field environment, such as... Figure 1 As shown.
[0049] A method for identifying the target attributes of dynamic targets by an unmanned vehicle in a field environment, the method comprising:
[0050] Step 101: The command and control terminal collects situational information in real time and sends it to the unmanned vehicle. The situational information includes the coordinates and attribute information of each target.
[0051] The command and control terminal refers to a remote control center located outside the unmanned vehicle. It integrates data acquired from various channels, such as satellites, radar, cameras, and infrared detectors. Situational information includes the coordinates and attribute information of each target. Coordinates are latitude, longitude, and altitude coordinates. Attribute information includes the target's type and characteristics; types include vehicles, personnel, and aircraft, while characteristics include size, color, and friend-or-foe information. Attribute information can be obtained through satellite remote sensing or radar technology, such as imaging ground targets using high-resolution optical cameras. By analyzing these images, the target's appearance features, such as shape, size, and color, can be extracted. Situational information is transmitted to the unmanned vehicle via a wireless communication network.
[0052] Step 102: The unmanned vehicle receives situational information, forms situational points based on the situational information and the time of reception, and generates a situational sequence by generating all situational points of the same target within a preset time period.
[0053] The preset time period refers to the length of a pre-defined time window, such as 10 seconds to 5 minutes, used to define the time range of situation points. After receiving situation information, the unmanned vehicle will form situation points based on this information and the time of reception. All situation points of the same target within the preset time period are integrated to generate a situation sequence.
[0054] Step 103: After the unmanned vehicle detects any target, it acquires the reconnaissance information of the target, including the reconnaissance coordinates and the reconnaissance time.
[0055] Autonomous vehicles use onboard sensors, such as cameras, radar, or lidar, to detect moving objects in the environment. Once a target is detected, the vehicle determines its position using its GPS or inertial navigation system, calculates the target's polar coordinates relative to the vehicle using radar ranging and angle data, and then performs coordinate transformation to obtain the target's detection coordinates. The detection time is the moment the autonomous vehicle detects the target.
[0056] Step 104: Determine whether the number of situation points in the situation sequence of each target meets a preset condition. If the preset condition is met, perform trajectory prediction for each target based on the situation sequence of the target to obtain the predicted coordinates of each target at the reconnaissance time. The preset condition is: the number of situation points in the situation sequence of the target is greater than or equal to a situation point threshold.
[0057] When the number of situation points is greater than or equal to the situation point threshold, trajectory prediction is performed for each target based on the target's situation sequence. If the number of situation points is less than the situation point threshold, the coordinates of the last situation point in each target's situation sequence are used as the predicted coordinates of that target at the reconnaissance time. The situation point threshold is 10, meaning the preset condition is that the number of situation points in the situation sequence is greater than or equal to 10.
[0058] If the number of situation points is less than 10, the accuracy of trajectory prediction by Kalman filtering is low due to the limited number of situation points. In this case, the coordinates of the last situation point in each target situation sequence are used as the predicted coordinates of the target at the reconnaissance time.
[0059] If the number of situation points in the target's situation sequence meets a preset threshold, then the Kalman filter algorithm is used to predict the target's trajectory, obtaining the target's predicted coordinates at the reconnaissance time. The specific steps include:
[0060] Step 1041: Input all situation points of the situation sequence of targets that are greater than or equal to the situation point threshold into the initial Kalman filter model, and update the Kalman gain of the initial Kalman filter model based on the situation sequence to obtain the final Kalman model.
[0061] The final Kalman model is obtained by updating the Kalman gain of the initial Kalman filter model based on the situation sequence, including:
[0062] Starting from the second situation point, perform the following operations sequentially for each situation point in the situation sequence:
[0063] Step 10411: Input the coordinates of the situation point at the previous time step and the receiving time into the Kalman filter model to calculate the estimated coordinates at the current time step;
[0064] The coordinates of the previous moment and its reception time are input into the Kalman filter model, which calculates the estimated coordinates for the current moment. The Kalman filter model dynamically adjusts the Kalman gain parameter based on each observed coordinate estimate, gradually approximating the target's actual trajectory. Calculating the estimated coordinates for the current moment requires using the coordinates of the situation point at the current and previous moments, as well as the reception time interval of the situation point, to calculate the target's velocity and turning angle at the previous moment. Based on the velocity, turning angle, reception time interval, and current Kalman gain, the estimated coordinates for the current moment are calculated.
[0065] Step 10412: The coordinates of the current situation point are used as coordinate observations, and the Kalman gain is updated based on the coordinate observations and the estimated coordinates. The Kalman gain is used to measure the impact of the difference between the observed and estimated values on the model parameters. The coordinates of the current situation point are compared with the estimated coordinates at the current time, the difference between the two is calculated, and the Kalman gain is updated based on this difference.
[0066] The Kalman gain update formula is:
[0067] α( )
[0068] -
[0069] -
[0070] in, This represents the difference between the x-axis coordinates of the current situation point and its estimated coordinates at the current moment. Let y be the difference between the current position point's coordinates and the estimated coordinates at the current moment, and α be the preset learning rate. For the Kalman gain from the last update, The Kalman gain updated at the current moment. This is the estimated x-axis coordinate at the current moment. This is the estimated y-axis coordinate at the current moment. The x-coordinate of the current situation point. The y-coordinate of the current situation point.
[0071] Step 10413: Use the last updated Kalman gain as the final gain of the Kalman filter model to obtain the final Kalman model.
[0072] Step 1042: Input the situation information of the last situation point in the situation sequence and the time interval between the reception time and the reconnaissance time of the last situation point into the final Kalman model to obtain the coordinates of the target at the reconnaissance time.
[0073] Input the situation information of the last situation point in the situation sequence and the time interval between the reception time and the reconnaissance time of the last situation point into the final Kalman model to obtain the target's coordinates at the reconnaissance time as shown in the following formula;
[0074]
[0075]
[0076] in, , The coordinates of the target at the time of reconnaissance. , The coordinates of the last situation point. The target velocity at the moment of reception of the last situation point. The turning angle at the moment of receiving the last situation point. The time interval between the reception time of the last situation point and the reconnaissance time. For Kalman gain.
[0077] The target velocity at the last moment of situational awareness reception is shown in the following formula;
[0078] ;
[0079] ;
[0080] ;
[0081] in, The x-axis component of the target velocity at the last state point reception time. The target velocity y-axis component is the value received at the last state point. The time interval between the last situation point and the previous situation point. , Set the coordinates of the last state point at the previous moment;
[0082] The turning angle at the moment of receiving the last situation point is shown in the following formula;
[0083] .
[0084] Before obtaining the final Kalman model by updating the Kalman gain of the initial Kalman filter model based on the situation sequence, the following steps are also included:
[0085] Convert the coordinates of situation points in the situation sequence to rectangular coordinates. The coordinates of situation points are latitude, longitude, and height coordinates, which can be converted to plane rectangular coordinates using Mercator projection, Gauss-Kruger projection, etc.
[0086] Step 105: Match the predicted coordinates of all targets at the reconnaissance time with the reconnaissance coordinates to obtain the matched predicted coordinates, and use the attribute information of the target corresponding to the matched predicted coordinates as the attribute information of the reconnaissance target.
[0087] Matching the predicted coordinates of all targets at the reconnaissance time with the reconnaissance coordinates includes:
[0088] Calculate the distance between the predicted coordinates of all targets at the reconnaissance time and the reconnaissance coordinates;
[0089] The target with the shortest distance between the predicted coordinates at the reconnaissance time and the reconnaissance coordinates is selected as the matching target.
[0090] The spatial distance between the predicted and actual detected positions of all targets at the reconnaissance time is calculated, and the target with the smallest distance is selected as the matching result. When calculating the distance, all predicted coordinates are traversed, and the straight-line distance between each predicted coordinate and the reconnaissance coordinate on the plane is calculated. The distance between the predicted and reconnaissance coordinates at the reconnaissance time is the Euclidean distance, which quantifies the spatial deviation between the predicted and actual positions. By comparing all calculation results, the target with the smallest value is selected as the matching target. If multiple targets have the same distance value, additional attributes of the targets are compared, such as the closest velocity or the closest direction.
[0091] The attribute information of the reconnaissance target determined based on the matching results includes:
[0092] The attribute information of the matching target is assigned to the reconnaissance target, thereby obtaining the attribute information of the reconnaissance target.
[0093] The above embodiments of the present invention include at least the following beneficial effects: The present invention receives situational information sent by the command and control terminal of an unmanned vehicle and establishes a situational sequence. It uses the Kalman filter algorithm to predict the trajectory of the target and matches it with the reconnaissance coordinates at the reconnaissance time. This enables accurate identification of the attribute information of dynamic targets, fully considering the dynamic characteristics of the targets. Through prediction and matching, the accuracy of target attribute identification is improved. Based on the situational sequence, the present invention effectively solves the problem of mismatch between situational target information and actual reconnaissance target information caused by time delay by introducing preset conditions and situational point thresholds. When the number of situational points does not meet the preset conditions, the coordinates of the last situational point are used as the prediction coordinates, reducing the impact of time delay on target attribute identification. When the number of situational points in the target's situational sequence meets the preset conditions, the Kalman filter algorithm is used for trajectory prediction, and the prediction accuracy of the Kalman filter model is further improved by converting the situational point coordinates in the situational sequence to rectangular coordinates.
[0094] Those skilled in the art will understand that all or part of the processes of the methods described in the above embodiments can be implemented by a computer program instructing related hardware, and the program can be stored in a computer-readable storage medium. The computer-readable storage medium may be a disk, optical disk, read-only memory, or random access memory, etc.
[0095] The above description is only a preferred embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any changes or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in the present invention should be included within the scope of protection of the present invention.
Claims
1. A method for identifying the target attributes of dynamic targets by an unmanned vehicle in a field environment, characterized in that, The method includes: The command and control terminal collects situational information in real time and sends it to the unmanned vehicle. The situational information includes the coordinates and attribute information of each target. The autonomous vehicle receives situational information and forms situational points based on the situational information and the time of reception. It then generates a situational sequence by combining all situational points of the same target within a preset time period. After the unmanned vehicle detects any target, it acquires the reconnaissance information of the target, including the reconnaissance coordinates and the reconnaissance time. Determine whether the number of situation points in the situation sequence of each target meets the preset conditions. When the preset conditions are met, perform trajectory prediction for each target based on the situation sequence of the target to obtain the predicted coordinates of each target at the reconnaissance time. Among them, the trajectory prediction for each target based on the situation sequence of the target uses the Kalman filter algorithm. The target velocity at the time of receiving the last situation point of the situation sequence of each target is shown in the following formula. ； ； ； in, The x-axis component of the target velocity at the last state point reception time. The target velocity y-axis component is the value received at the last state point. The time interval between the last situation point and the previous situation point. , Set the coordinates of the last state point at the previous moment; The turning angle at the moment of receiving the last situation point is shown in the following formula; ； The predicted coordinates of all targets at the reconnaissance time are matched with the reconnaissance coordinates to obtain the matched predicted coordinates. The attribute information of the target corresponding to the matched predicted coordinates is used as the attribute information of the reconnaissance target.
2. The method according to claim 1, characterized in that, The preset condition is: the number of situation points in the target's situation sequence is greater than or equal to the situation point threshold. If the preset conditions are not met, the coordinates of the last situation point in each target situation sequence will be used as the predicted coordinates of the target at the reconnaissance time.
3. The method according to claim 1, characterized in that, Matching the predicted coordinates of all targets at the reconnaissance time with the reconnaissance coordinates includes: Calculate the distance between the predicted coordinates of all targets at the reconnaissance time and the reconnaissance coordinates; The target with the shortest distance between the predicted coordinates at the reconnaissance time and the reconnaissance coordinates is selected as the matching target.
4. The method according to claim 3, characterized in that, The attribute information of the reconnaissance target determined based on the matching results includes: The attribute information of the matching target is assigned to the reconnaissance target, thereby obtaining the attribute information of the reconnaissance target.
5. The method according to claim 2, characterized in that, If the number of situation points in the target's situation sequence meets a preset threshold, then the Kalman filter algorithm is used to predict the target's trajectory, obtaining the target's predicted coordinates at the reconnaissance time. The specific steps include: Input all situation points of the situation sequence of targets greater than or equal to the situation point threshold into the initial Kalman filter model, and update the Kalman gain of the initial Kalman filter model based on the situation sequence to obtain the final Kalman model; The situation information of the last situation point in the situation sequence, as well as the time interval between the reception time and the reconnaissance time of the last situation point, are input into the final Kalman model to obtain the target's coordinates at the reconnaissance time.
6. The method according to claim 5, characterized in that, The final Kalman model is obtained by updating the Kalman gain of the initial Kalman filter model based on the situation sequence, including: Starting from the second situation point, perform the following operations sequentially for each situation point in the situation sequence: Input the coordinates of the previous situation point and the receiving time into the Kalman filter model to calculate the estimated coordinates at the current time. The coordinates of the current situation point are used as coordinate observations, and the Kalman gain is updated based on the coordinate observations and the estimated coordinates. The final Kalman model is obtained by using the last updated Kalman gain as the final gain of the Kalman filter model.
7. The method according to claim 6, characterized in that, Input the situation information of the last situation point in the situation sequence and the time interval between the reception time and the reconnaissance time of the last situation point into the final Kalman model to obtain the target's coordinates at the reconnaissance time as shown in the following formula; in, , The coordinates of the target at the time of reconnaissance. , The coordinates of the last situation point. The target velocity at the moment of reception of the last situation point. The turning angle at the moment of receiving the last situation point. The time interval between the reception time of the last situation point and the reconnaissance time. This is the Kalman gain.
8. The method according to claim 7, characterized in that, The situation point coordinates in the situation sequence are latitude, longitude, and altitude coordinates; before updating the Kalman gain of the initial Kalman filter model based on the situation sequence to obtain the final Kalman model, the process also includes: Convert the coordinates of the situation points in the situation sequence to rectangular coordinates.
9. The method according to claim 2, characterized in that, The threshold for the situation point is 10.