Optoelectronic multi-target track processing method based on route matching and random finite set

Abstract

The application discloses an optical-electric multi-target track processing method based on a route matching and a random finite set, mainly extracts target information in a field of view of an optical-electric sensor, and screens out a cross-section target measurement set; uses the random finite set to generate a conventional target track set and a newly updated cross-section target track set at a current moment; uses a cross-section target track at a previous moment to match a most similar route in a pre-constructed route matching library; and uses the most similar route to predict a bearing value of the target at the current moment, finds a closest measurement value in the cross-section target measurement set at the current moment, and uses the measurement value to update the track at the previous moment; and the updated cross-section target track set is combined with the cross-section target track set. The application can realize the association of measurement values and tracks of multiple targets, and realizes route construction of a cross-section fast-moving target.

Description

Technical Field

[0001] This invention relates to the field of trajectory processing, and more particularly to a photoelectric multi-target trajectory processing method and system based on route matching and random finite sets. Background Technology

[0002] A photoelectric sensor is a purely orientation-detecting sensor; it can only acquire the target's azimuth and elevation information, but not its distance information. Using only the target's azimuth and elevation information, it is impossible to directly calculate the target's accurate position. Therefore, navigation based on photoelectric sensor measurements is extremely difficult.

[0003] A photoelectric sensor is also an image sensor; for the same target, as the distance between the target and the sensor decreases, the image of the target in the photoelectric sensor also increases. For example... Figure 1 As shown, if the photoelectric sensor is considered as a point mass, the perpendicular distance between this point mass and the target's flight path is called the intercept. For a fast-moving target that crosses the intercept, the closer the target is to the intercept, the larger the target size will be, and the greater the rate of change of the target's azimuth angular velocity detected by the photoelectric sensor will be, making it difficult to establish navigation for the target.

[0004] Currently, many scholars often use Bayesian estimation methods such as Kalman filtering to estimate the target's position information based on its angle measurements, thereby achieving target navigation. However, this method is often only effective for single targets. When multiple targets exist within the observation range of the photoelectric sensor, this method cannot correlate the measured values ​​with the flight path, thus failing to achieve target navigation for multiple targets. Furthermore, because the rate of change of the target angle measured by the sensor is particularly large when the target crosses the yoke, it is difficult to achieve target navigation using Bayesian estimation methods in such cases.

[0005] For multi-target navigation, random finite sets have been widely used in recent years. Random finite sets can correlate measurements with trajectories, but this method is also difficult to implement for navigation of fast-moving targets with cross-sections. Summary of the Invention

[0006] This invention addresses the problem of difficulty in establishing navigation for targets using target measurement information from photoelectric sensors, and provides a method and system that can achieve stable navigation for slow-moving targets, fast-moving targets, and targets that cross the diameter.

[0007] The technical solution adopted in this invention is:

[0008] A method for processing photoelectric multi-target tracks based on route matching and random finite sets is provided, including the following steps:

[0009] S1. Construct a route matching library for the target, in which each route corresponds to the motion characteristics of a target with a cross-section;

[0010] S2. Initialize the target's size threshold and grayscale threshold;

[0011] S3. Extract target information within the field of view of the photoelectric sensor, including the pixel coordinates of the target in the image coordinate system, the size of the target, and the average gray value;

[0012] S4. Using the inertial navigation information from the photoelectric sensor, the pixel coordinates of the target are converted into azimuth and pitch information in the geographic system.

[0013] S5. Select the target information whose average gray value and size are both greater than the corresponding threshold as the over-cut target measurement set, and the remaining target information as the regular target measurement set.

[0014] S6. Using the set of regular target measurements, the set of regular target tracks from the previous time step, and the set of cross-cut target measurements from the current time step, generate the set of regular target tracks and the newly updated set of cross-cut target tracks from the current time step using a random finite set.

[0015] S7. Use the target track with the cross-cutting path from the previous moment to find the most similar route in the route matching library;

[0016] S8. Use the most similar route to predict the target's bearing value at the current time. Find the measurement value that is closest to the target's bearing prediction value at the current time in the target measurement set at the current time, and use the measurement value to update the track at the previous time to obtain the target track set at the current time.

[0017] S9. Merge the updated cross-cut target track set with the cross-cut target track set in step S6.

[0018] Following the above technical solution, the specific steps for constructing the target route matching library are as follows:

[0019] The target is periodically measured by photoelectric sensors to obtain all the measurement values ​​of each target from the start to the cross-section, forming a set of measurement information for each target at a given speed. The motion curve of the target is fitted by the least squares method, and each motion curve corresponds to a route. A route matching library is constructed based on multiple routes.

[0020] Following the above technical solution, step S6 specifically involves:

[0021] The current set of cross-cut target measurements, the set of regular target measurements, and the set of regular target tracks from the previous time step are fed together into the multi-label Bernoulli random finite set algorithm.

[0022] Establish a matching relationship between the input track and the measurement value at the current time, and select the measurement value with a high matching degree to update the target track status information, including azimuth and pitch information;

[0023] In a random finite set, a track updated using cross-cut target measurements will evolve into a cross-cut target track, while a track updated using regular target measurements will be a regular target track. All cross-cut target tracks constitute the cross-cut target track set at the current moment, and the tracks generated from the regular target track set constitute the regular target track set at the current moment.

[0024] Following the above technical solution, the random finite set algorithm is divided into prediction, grouping and updating, and Kalman filtering is used to predict and update the target trajectory state information.

[0025] Following the above technical solution, step S7 specifically includes:

[0026] 1) Select the azimuth values ​​of the target for any consecutive n sampling periods from the target measurement set with cross-section, and calculate the actual rate of change of the target's azimuth angle within these n sampling periods; n is a natural number;

[0027] 2) Calculate the estimated average rate of change of the target azimuth angle for each route in the route matching library within 5 sampling periods for each of the corresponding n sampling periods;

[0028] 3) Calculate the error between the estimated average rate of change of the target azimuth angle obtained for each route at the corresponding time in step 2) and the true value obtained in step 1). This error is called the matching error.

[0029] 4) Select the route with the smallest matching error as the final matching result.

[0030] Following the above technical solution, n≥5.

[0031] This invention also provides an optoelectronic multi-target trajectory processing system based on route matching and random finite sets, comprising:

[0032] The route matching library construction module is used to build a route matching library for targets. Each route in the route matching library corresponds to the motion characteristics of a target with a cross-section.

[0033] The initialization module is used to initialize the target's size threshold and grayscale threshold.

[0034] The target information extraction module is used to extract target information within the field of view of the photoelectric sensor, including the pixel coordinates of the target in the image coordinate system, the size of the target, and the average gray value; and to use the inertial navigation information of the photoelectric sensor to convert the pixel coordinates of the target into azimuth and pitch information in the geographic system.

[0035] The filtering module is used to filter out target information whose average gray value and size are both greater than the corresponding threshold as the over-cut target measurement set, and the remaining target information as the regular target measurement set.

[0036] The target trajectory generation module is used to generate the current set of regular target trajectories and the newly updated set of cross-cut target trajectories using a random finite set, by utilizing the regular target measurement set, the regular target trajectory set of the previous time step, and the cross-cut target measurement set of the current time step.

[0037] The route matching module is used to match the most similar route in the route matching library using the target track with the cross-cutting path from the previous moment.

[0038] The track update module is used to predict the bearing value of the target at the current time using the most similar matching route, find the measurement value that is closest to the predicted bearing value of the target at the current time in the cross-section target measurement set at the current time, and use the measurement value to update the track at the previous time to obtain the cross-section target track set at the current time.

[0039] The track merging module is used to merge the updated cross-cut target track set with the cross-cut target track set generated in the target track generation module.

[0040] Following the above technical solution, the route matching library construction module is specifically used for:

[0041] The target is periodically measured by photoelectric sensors to obtain all the measurement values ​​of each target from the start to the cross-section, forming a set of measurement information for each target at a given speed. The motion curve of the target is fitted by the least squares method, and each motion curve corresponds to a route. A route matching library is constructed based on multiple routes.

[0042] Following the above technical solution, the target trajectory generation module is specifically used for:

[0043] The current set of cross-cut target measurements, the set of regular target measurements, and the set of regular target tracks from the previous time step are fed together into the multi-label Bernoulli random finite set algorithm.

[0044] Establish a matching relationship between the input track and the measurement value at the current time, and select the measurement value with a high matching degree to update the target track status information, including azimuth and pitch information;

[0045] In a random finite set, a track updated using cross-cut target measurements will evolve into a cross-cut target track, while a track updated using regular target measurements will be a regular target track. All cross-cut target tracks constitute the cross-cut target track set at the current moment, and the tracks generated from the regular target track set constitute the regular target track set at the current moment.

[0046] The present invention also provides a computer storage medium storing a computer program executable by a processor, the computer program executing the photoelectric multi-target trajectory processing method based on route matching and random finite sets described in the above technical solution.

[0047] The beneficial effects of this invention are as follows: Addressing the problem that conventional random finite set algorithms cannot achieve stable navigation for targets that rapidly cross the circumference, this invention constructs a route matching library based on the motion characteristics of the target. It uses the target's motion characteristics over a period of time to match routes in the route matching library, and then uses the matched routes instead of directly using random finite set algorithms to predict the target's bearing value at the current moment. This achieves stable navigation for both conventionally maneuvering targets and targets that rapidly maneuver across the circumference. Attached Figure Description

[0048] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0049] Figure 1 This is a schematic diagram of the target through-cut diameter in an embodiment of the present invention;

[0050] Figure 2 This is a schematic diagram of the algorithm structure of the photoelectric multi-target trajectory processing method based on route matching and random finite sets in an embodiment of the present invention;

[0051] Figure 3 This is a flowchart of the photoelectric multi-target trajectory processing method based on route matching and random finite sets according to an embodiment of the present invention;

[0052] Figure 4 This is a flowchart of the random finite set algorithm according to an embodiment of the present invention;

[0053] Figure 5 This is a flowchart of another embodiment of the present invention, which is a photoelectric multi-target trajectory processing method based on route matching and random finite sets. Detailed Implementation

[0054] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the invention.

[0055] Example 1

[0056] like Figure 5As shown, this embodiment of the photoelectric multi-target trajectory processing method based on route matching and random finite sets includes the following steps:

[0057] S1. Construct a route matching library for the target, in which each route corresponds to the motion characteristics of a target with a cross-section;

[0058] S2. Initialize the target's size threshold and grayscale threshold;

[0059] S3. Extract target information within the field of view of the photoelectric sensor, including the pixel coordinates of the target in the image coordinate system, the size of the target, and the average gray value;

[0060] S4. Using the inertial navigation information from the photoelectric sensor, the pixel coordinates of the target are converted into azimuth and pitch information in the geographic system.

[0061] S5. Select the target information whose average gray value and size are both greater than the corresponding threshold as the over-cut target measurement set, and the remaining target information as the regular target measurement set.

[0062] S6. Using the set of conventional target measurements, the set of conventional target tracks from the previous time step, and the set of cross-sectional target measurements from the current time step, generate the set of conventional target tracks and the newly updated set of cross-sectional target tracks from the current time step using a random finite set. Specifically, the track updated using the measurement values ​​from the set of conventional target measurements from the previous time step is defined as a conventional target track, and all conventional target tracks together form the set of conventional target tracks. Similarly, the track updated using the measurement values ​​from the set of cross-sectional target measurements from the previous time step is defined as a cross-sectional target track, and all cross-sectional target tracks together form the set of cross-sectional target tracks.

[0063] S7. Use the target track with the cross-cutting path from the previous moment to find the most similar route in the route matching library;

[0064] S8. Use the most similar route to predict the target's bearing value at the current time. Find the measurement value that is closest to the target's bearing prediction value at the current time in the target measurement set at the current time, and use the measurement value to update the track at the previous time to obtain the target track set at the current time.

[0065] S9. Merge the updated cross-cut target track set with the cross-cut target track set in step S6.

[0066] Furthermore, the construction of the target route matching library is specifically as follows: the target is periodically measured by photoelectric sensors to obtain all the measurement values ​​of each target from the start to the cross-section, forming a set of measurement information of each target at this given speed. The motion curve of the target is fitted by the least squares method, and each motion curve corresponds to a route. The route matching library is constructed based on multiple routes.

[0067] Specifically, step S6 is as follows:

[0068] The current set of cross-cut target measurements, the set of regular target measurements, and the set of regular target tracks from the previous time step are fed together into the multi-label Bernoulli random finite set algorithm.

[0069] Establish a matching relationship between the input track and the measurement value at the current time, and select the measurement value with a high matching degree to update the target track status information, including azimuth and pitch information;

[0070] In a random finite set, a track updated using cross-cut target measurements will evolve into a cross-cut target track, while a track updated using regular target measurements will be a regular target track. All cross-cut target tracks constitute the cross-cut target track set at the current moment, and the tracks generated from the regular target track set constitute the regular target track set at the current moment.

[0071] In this embodiment, the random finite set algorithm is divided into prediction, grouping, and updating, and Kalman filtering is used to predict and update the target track state information.

[0072] Furthermore, step S7 specifically includes:

[0073] 1) Select the azimuth values ​​of the target for any consecutive n sampling periods from the target measurement set with cross-section, and calculate the actual rate of change of the target's azimuth angle within these n sampling periods; n is a natural number;

[0074] 2) Calculate the estimated average rate of change of the target azimuth angle for each route in the route matching library within 5 sampling periods for each of the corresponding n sampling periods;

[0075] 3) Calculate the error between the estimated average rate of change of the target azimuth angle obtained for each route at the corresponding time in step 2) and the true value obtained in step 1). This error is called the matching error.

[0076] 4) Select the route with the smallest matching error as the final matching result.

[0077] A photoelectric multi-target trajectory processing method based on route matching and random finite sets is characterized in that n≥5.

[0078] This invention, based on route matching and random finite sets, enables the correlation between measurement values ​​and flight paths even when multiple targets exist within the observation range of a photoelectric sensor. It achieves stable flight path establishment for targets with conventional maneuvers and also for targets with rapid maneuvers exceeding their cross-section.

[0079] Example 2

[0080] This embodiment is based on embodiment 1, but the difference is that it further provides the specific implementation methods used in each step.

[0081] In this embodiment, such as Figure 2 and 3 As shown, the photoelectric multi-target trajectory processing method based on route matching and random finite sets includes the following steps:

[0082] (1) A route matching library is generated by fitting a ninth-order polynomial (see Formula 5) to describe the motion characteristics of targets that cross the circumference.

[0083] Assume the geographic coordinates of the photoelectric sensor at time t are... The geographic coordinates of the target at time t

[0084] The sensor target orientation measurement model is as follows:

[0085]

[0086] The motion model of a target moving at a constant velocity in a straight line through the cross section is as follows:

[0087] X(t+1)=FX(t) (2)

[0088]

[0089]

[0090] Where: T is the sampling interval. and Let be the coordinates of the target at time t in the geographic system. and Let x and y be the velocities of the target at time t in the geographic frame, respectively.

[0091] Choose several different speeds, that is, choose different and For targets moving at a constant velocity in a straight line across different cross-sections, a sensor orientation measurement model is used to measure the targets at a sampling period of T. This obtains all measurement values ​​for each target from the start to the point of crossing the cross-section. These discrete measurement values ​​constitute the measurement set of that target at a given velocity. Using this discrete measurement set The least squares method is used to fit the target's motion curve. The fitted motion curve is the target's path at a given speed. The set of motion curves constitutes a path matching library.

[0092] The expression for each route in the route matching library is as follows:

[0093]

[0094] Where: p i Let be the coefficient of the term of degree i, which is obtained by least squares fitting. The specific implementation of the fitting is as follows:

[0095] Select azimuth measurement values ​​from the measurement set The data source set is composed of the relative time t. The fitting formula is as follows:

[0096] A = (M T M) -1 M T Z (6)

[0097] in:

[0098]

[0099] A = [p1 p2 ... p9] T (8)

[0100] Z = [az1 az2 ... az] n ] T (9)

[0101] (2) Initialize the size threshold BS lim Gray threshold lim If both the size and grayscale of the target are greater than the threshold, the target may be an over-cut target.

[0102] (3) Use common target extraction algorithms (such as the TopHat algorithm) to extract target information within the field of view of the photoelectric sensor, including the target size tg. size The average gray value tg of the target gray The pixel coordinates (posx,posy) of the target in the image coordinate system.

[0103] (4) Using the inertial navigation information from the photoelectric sensor, the pixel coordinates of the target are converted into its azimuth in the geographic system. tg and pitch information tg The conversion method is as follows:

[0104]

[0105] Among them: az sensor With pitch sensor This refers to the orientation of the photoelectric sensor, information that can be obtained through an inertial navigation device paired with the photoelectric sensor, resolX. sensor and resolY sensor These are the sensor's angular resolutions, calculated as follows:

[0106]

[0107] Among them: FOV X This represents the sensor's lateral field of view. (resolX) img FOV (Field of View) is the number of pixels in the horizontal direction of the image output by the sensor. Y This represents the sensor's lateral field of view. (resolY) img This represents the number of pixels in the vertical direction of the sensor output image.

[0108] (5) The measurement information set in step (3) where the target size is greater than the size threshold in step (2) and the target average gray level is greater than the gray level threshold in step (2) is taken as the cross-section target measurement set. The remaining measurement information set is taken as the regular target measurement set.

[0109] (6) Set the current cross-section target measurement set Θ S (t), the set of conventional target measurements Θ N (t), and the set of regular target tracks Ξ from the previous time step. N (t-1) are fed together into the multi-label Bernoulli random finite set algorithm. For example... Figure 4 As shown, the random finite set algorithm consists of several main steps: prediction, grouping, and updating. Kalman filtering is used for the prediction and updating of the target state within the random finite set. Since the random finite set algorithm takes many measurements as input simultaneously, it first needs to establish a matching relationship between the input track and the measurements at the current moment. The random finite set selects measurements with high matching degrees and uses the Kalman algorithm to update the target track state information (azimuth, pitch). If the target measurement value Θ with an over-cutoff is used in the random finite set... S (t) The updated track will evolve into a cross-cut target track if conventional target measurement Θ is used. N (t) The updated track is the regular target track. All cross-cutting target tracks constitute the current cross-cutting target track set Ξ. 1 S (t), the tracks generated by the set of regular target tracks constitute the set of regular target tracks Ξ at the current moment. 1 N (t). Combining route matching with random finite sets can compensate for the poor performance of random finite sets on fast targets with large cross-sections.

[0110] (7) Set the target tracks from the previous time step Ξ S (t-1) and the cross-section target measurement set Θ at the current time. S (t-1) Perform route matching. The specific process for route matching is as follows:

[0111] 1) Randomly select the azimuth values ​​of the target for N consecutive sampling periods (N≥5 in this invention, N is 5 in this embodiment) from the target measurement set of the cross-section target, and calculate the actual azimuth angle change rate of the target within these 5 sampling periods.

[0112] 2) Calculate the estimated average rate of change of the target azimuth angle for each route within the corresponding 5 sampling periods.

[0113] 3) Calculate the error between the estimated average rate of change of the target azimuth angle obtained for each route at the corresponding time in step 2) and the actual value obtained in step 1). This error is called the matching error.

[0114] 4) Select the route with the smallest matching error as the final matching result.

[0115] (8) Use the matched route to predict the azimuth value of the target at the current time, then find the measurement value that is closest to the predicted azimuth value of the target at the current time in the cross-cut target measurement set at the current time, and use the measurement value to update the track at the previous time to obtain the cross-cut target track set at the current time.

[0116] (9) Merge the target track set with the target track set with the circumference in step (8) with the target track set with the circumference in step (6).

[0117] Example 3

[0118] This embodiment is mainly used to implement method embodiment 1. The photoelectric multi-target trajectory processing system based on route matching and random finite sets includes:

[0119] The route matching library construction module is used to build a route matching library for targets. Each route in the route matching library corresponds to the motion characteristics of a target with a cross-section.

[0120] The initialization module is used to initialize the target's size threshold and grayscale threshold.

[0121] The target information extraction module is used to extract target information within the field of view of the photoelectric sensor, including the pixel coordinates of the target in the image coordinate system, the size of the target, and the average gray value; and to use the inertial navigation information of the photoelectric sensor to convert the pixel coordinates of the target into azimuth and pitch information in the geographic system.

[0122] The filtering module is used to filter out target information whose average gray value and size are both greater than the corresponding threshold as the over-cut target measurement set, and the remaining target information as the regular target measurement set.

[0123] The target trajectory generation module is used to generate the current set of regular target trajectories and the newly updated set of cross-cut target trajectories using a random finite set, by utilizing the regular target measurement set, the regular target trajectory set of the previous time step, and the cross-cut target measurement set of the current time step.

[0124] The route matching module is used to match the most similar route in the route matching library using the target track with the cross-cutting path from the previous moment.

[0125] The track update module is used to predict the bearing value of the target at the current time using the most similar matching route, find the measurement value that is closest to the predicted bearing value of the target at the current time in the cross-section target measurement set at the current time, and use the measurement value to update the track at the previous time to obtain the cross-section target track set at the current time.

[0126] The track merging module is used to merge the updated cross-cut target track set with the cross-cut target track set generated in the target track generation module.

[0127] Specifically, the route matching library construction module is used for:

[0128] The target is periodically measured by photoelectric sensors to obtain all the measurement values ​​of each target from the start to the cross-section, forming a set of measurement information for each target at a given speed. The motion curve of the target is fitted by the least squares method, and each motion curve corresponds to a route. A route matching library is constructed based on multiple routes.

[0129] The target trajectory generation module is specifically used for:

[0130] The current set of cross-cut target measurements, the set of regular target measurements, and the set of regular target tracks from the previous time step are fed together into the multi-label Bernoulli random finite set algorithm.

[0131] Establish a matching relationship between the input track and the measurement value at the current time, and select the measurement value with a high matching degree to update the target track status information, including azimuth and pitch information;

[0132] In a random finite set, a track updated using cross-cut target measurements will evolve into a cross-cut target track, while a track updated using regular target measurements will be a regular target track. All cross-cut target tracks constitute the cross-cut target track set at the current moment, and the tracks generated from the regular target track set constitute the regular target track set at the current moment.

[0133] The modules in this system embodiment are mainly used to implement the various steps of the method embodiment, and will not be described in detail here.

[0134] Example 4

[0135] This application also provides a computer-readable storage medium, such as flash memory, hard disk, multimedia card, card-type memory (e.g., SD or DX memory), random access memory (RAM), static random access memory (SRAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), programmable read-only memory (PROM), magnetic memory, disk, optical disk, server, App application store, etc., which stores a computer program. When the program is executed by a processor, it implements the corresponding function. In this embodiment, the computer-readable storage medium, when executed by a processor, implements the photoelectric multi-target trajectory processing method based on route matching and random finite sets, as described in the method embodiment.

[0136] It should be noted that, depending on the implementation needs, the various steps / components described in this application can be broken down into more steps / components, or two or more steps / components or parts of the operation of steps / components can be combined into new steps / components to achieve the purpose of this invention.

[0137] The order of the steps in the above embodiments does not imply the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of this application.

[0138] It should be understood that those skilled in the art can make improvements or modifications based on the above description, and all such improvements and modifications should fall within the protection scope of the appended claims.

Claims

1. A photoelectric multi-target track processing method based on route matching and random finite set, characterized in that, Includes the following steps: S1. Construct a route matching library for the target, in which each route corresponds to the motion characteristics of a target with a cross-section; S2. Initialize the target's size threshold and grayscale threshold; S3. Extract target information within the field of view of the photoelectric sensor, including the pixel coordinates of the target in the image coordinate system, the size of the target, and the average gray value; S4. Using the inertial navigation information from the photoelectric sensor, the pixel coordinates of the target are converted into azimuth and pitch information in the geographic system. S5. Select the target information whose average gray value and size are both greater than the corresponding threshold as the over-cut target measurement set, and the remaining target information as the regular target measurement set. S6. Using the set of regular target measurements, the set of regular target tracks from the previous time step, and the set of cross-cut target measurements from the current time step, generate the set of regular target tracks and the newly updated set of cross-cut target tracks from the current time step using a random finite set. S7. Use the target track with the cross-cutting path from the previous moment to find the most similar route in the route matching library; S8. Use the most similar route to predict the target's bearing value at the current time. Find the measurement value that is closest to the target's bearing prediction value at the current time in the target measurement set at the current time, and use the measurement value to update the track at the previous time to obtain the target track set at the current time. S9. Merge the updated cross-cut target track set with the cross-cut target track set in step S6.

2. The photoelectric multi-target track processing method based on route matching and random finite set according to claim 1, characterized in that, The specific steps for constructing the target route matching library are as follows: The target is periodically measured by photoelectric sensors to obtain all the measurement values ​​of each target from the start to the cross-section, forming a set of measurement information for each target at a given speed. The motion curve of the target is fitted by the least squares method, and each motion curve corresponds to a route. A route matching library is constructed based on multiple routes. 3.The photoelectric multi-target track processing method based on route matching and random finite set according to claim 1, characterized in that, Step S6 is as follows: The current set of cross-cut target measurements, the set of regular target measurements, and the set of regular target tracks from the previous time step are fed together into the multi-label Bernoulli random finite set algorithm. Establish a matching relationship between the input track and the measurement value at the current time, and select the measurement value with a high matching degree to update the target track status information, including azimuth and pitch information; In a random finite set, a track updated using cross-cut target measurements will evolve into a cross-cut target track, while a track updated using regular target measurements will be a regular target track. All cross-cut target tracks constitute the cross-cut target track set at the current moment, and the tracks generated from the regular target track set constitute the regular target track set at the current moment.

4. The photoelectric multi-target track processing method based on route matching and random finite set according to claim 3, characterized in that, The random finite set algorithm is divided into prediction, grouping, and updating, and Kalman filtering is used to predict and update the target trajectory state information.

5. The photoelectric multi-target trajectory processing method based on route matching and random finite sets according to claim 1, characterized in that, Step S7 is as follows: 1) Select the azimuth values ​​of the target for any consecutive n sampling periods from the target measurement set with cross-section, and calculate the actual rate of change of the target's azimuth angle within these n sampling periods; n is a natural number; 2) Calculate the estimated average rate of change of the target azimuth angle for each route in the route matching library within 5 sampling periods for each of the corresponding n sampling periods; 3) Calculate the error between the estimated average rate of change of the target azimuth angle obtained for each route at the corresponding time in step 2) and the true value obtained in step 1). This error is called the matching error. 4) Select the route with the smallest matching error as the final matching result.

6. The photoelectric multi-target trajectory processing method based on route matching and random finite sets according to claim 5, characterized in that, n≥5。 7. A photoelectric multi-target trajectory processing system based on route matching and random finite sets, characterized in that, include: The route matching library construction module is used to build a route matching library for targets. Each route in the route matching library corresponds to the motion characteristics of a target with a cross-section. The initialization module is used to initialize the target's size threshold and grayscale threshold. The target information extraction module is used to extract target information within the field of view of the photoelectric sensor, including the pixel coordinates of the target in the image coordinate system, the size of the target, and the average gray value. It also utilizes the inertial navigation information from photoelectric sensors to convert the target's pixel coordinates into azimuth and elevation information in a geographic system. The filtering module is used to filter out target information whose average gray value and size are both greater than the corresponding threshold as the over-cut target measurement set, and the remaining target information as the regular target measurement set. The target trajectory generation module is used to generate the current set of regular target trajectories and the newly updated set of cross-cut target trajectories using a random finite set, by utilizing the regular target measurement set, the regular target trajectory set of the previous time step, and the cross-cut target measurement set of the current time step. The route matching module is used to match the most similar route in the route matching library using the target track with the cross-cutting path from the previous moment. The track update module is used to predict the bearing value of the target at the current time using the most similar matching route. It finds the measurement value that is closest to the predicted bearing value of the target at the current time in the cross-section target measurement set at the current time, and uses the measurement value to update the track at the previous time to obtain the cross-section target track set at the current time. The track merging module is used to merge the updated cross-cut target track set with the cross-cut target track set generated in the target track generation module.

8. The photoelectric multi-target trajectory processing system based on route matching and random finite sets according to claim 7, characterized in that, The route matching library construction module is specifically used for: The target is periodically measured by photoelectric sensors to obtain all the measurement values ​​of each target from the start to the cross-section, forming a set of measurement information for each target at a given speed. The motion curve of the target is fitted by the least squares method, and each motion curve corresponds to a route. A route matching library is constructed based on multiple routes.

9. The photoelectric multi-target trajectory processing system based on route matching and random finite sets according to claim 7, characterized in that, The target trajectory generation module is specifically used for: The current set of cross-cut target measurements, the set of regular target measurements, and the set of regular target tracks from the previous time step are fed together into the multi-label Bernoulli random finite set algorithm. Establish a matching relationship between the input track and the measurement value at the current time, and select the measurement value with a high matching degree to update the target track status information, including azimuth and pitch information; In a random finite set, a track updated using cross-cut target measurements will evolve into a cross-cut target track, while a track updated using regular target measurements will be a regular target track. All cross-cut target tracks constitute the cross-cut target track set at the current moment, and the tracks generated from the regular target track set constitute the regular target track set at the current moment.

10. A computer storage medium, characterized in that, It contains a computer program that can be executed by a processor, which performs the photoelectric multi-target trajectory processing method based on route matching and random finite sets as described in any one of claims 1-6.

Images