Target tracking method, device, equipment and medium based on line segment fitting

Abstract

The present invention discloses a target tracking method, device, equipment and medium based on line segment fitting, the method comprising: extracting several positions at the end of the target's motion trajectory and the corresponding time series, performing line segment fitting on the target, the line segment fitting comprising first-order straight line and second-order curve fitting; substituting the current position of the target and the corresponding time into the fitted first-order straight line model and second-order curve model to calculate the target's motion speed, motion direction, speed change and heading change under the first-order straight line model and second-order curve model respectively; assigning different weights to the first-order straight line model and the second-order curve model according to the speed change, size and heading change of the target and performing weighted averaging to obtain the final state information of the target. The present invention can realize target tracking for various motion states in complex scenes.

Description

Technical Field

[0001] The present invention belongs to the field of radar tracking technology, and in particular relates to a target tracking method, device, equipment and medium based on line segment fitting. Background Art

[0002] Radar can emit electromagnetic waves and receive target echoes to complete active detection of targets, and further establish the target's motion trajectory. The target tracking algorithm confirms the target and extracts information based on the target information reported by the radar front end, and provides stable, reliable, and accurate target information to the outside world. With the development of radar technology and the expansion of the market, radar's target detection capability is stronger, and the use scenarios and types of detection objects are more diverse and complex, which puts higher requirements on the target tracking algorithm.

[0003] The improvement of detection capability brings higher power and resolution, making the radar's detection of targets gradually change from "point" to "surface". The acquired radar images can obtain richer detailed information of the target, thereby achieving a clearer definition and classification of the target.

[0004] Due to the diversity of usage scenarios, radar can detect targets in various scenarios. First, there may be unfavorable factors such as multipath and occlusion in different areas of the scene, causing radar detection to have problems such as false alarms and missed alarms; second, the detection effect of radar varies for targets at different distances. The farther the distance, the more blurred the angle dimension detection effect and the lower the resolution. Although there are certain technical means to deal with these problems, it is still difficult for radar to achieve continuous and consistent detection effects in all areas.

[0005] The diversity of target objects, the movement characteristics of various types of targets are very different, with different maneuvers and accelerations. Even the same target may have different movement characteristics at different times.

[0006] The above new developments and changes bring new challenges to the original target tracking algorithms. Commonly used tracking algorithms include αβ, Kalman, UKF, and EKF. First, the algorithm performance is reduced for different detection errors at different distances and in different areas, for targets with different motion characteristics, or for sudden changes in the motion state of the same target, and a single state matrix or parameter model cannot meet the usage requirements. If state matrix switching or multiple models are used, the algorithm implementation complexity is greatly increased and the computational efficiency is reduced. Secondly, the Kalman algorithm occasionally has non-convergence problems during the iteration process. When the classic Kalman algorithm is used to process multiple sets of measured data, non-convergence problems occur, that is, the algorithm "loses lock and runs away" due to singular iteration matrices or other problems. According to the literature review, this type of problem is more serious in its derivative algorithms UKF and EKF. In some engineering practices, this is an unacceptable problem that must be solved. In order to achieve target tracking processing of two-dimensional radar target detection data, there is no solution that combines robustness, processing efficiency, and versatility. Summary of the invention

[0007] The purpose of the present invention is to overcome the defects of the prior art and provide a target tracking method, device, equipment and medium based on line segment fitting, which can achieve local optimization, ensure the accuracy and smoothness of the output motion trajectory, classify the target motion state according to the target attributes and motion parameters, and quickly and easily select and switch between multiple states according to the target motion state.

[0008] The object of the present invention is achieved through the following technical solutions:

[0009] A target tracking method based on line segment fitting, the method comprising:

[0010] Extracting several positions at the end of the target's motion trajectory and corresponding time series, and performing line segment fitting on the target, wherein the line segment fitting includes first-order straight line fitting and second-order curve fitting;

[0011] Substitute the current position of the target and the corresponding time into the fitted first-order straight line model and second-order curve model to calculate the target's movement speed, movement direction, speed change and heading change under the first-order straight line model and second-order curve model respectively;

[0012] Different weights are assigned to the first-order straight line model and the second-order curve model according to the speed change, size and heading change of the target, and weighted average is performed to obtain the final state information of the target.

[0013] Furthermore, performing first-order straight line and second-order curve fitting on the target specifically includes:

[0014] Extracting a plurality of positions at the end of the target motion trajectory and corresponding time series of the plurality of positions;

[0015] The first-order straight line fitting and the second-order curve fitting are performed on the X-axis and the Y-axis respectively.

[0016] Further, the step of substituting the fitted first-order straight line model and second-order curve model into the current position of the target and the corresponding time to calculate the movement speed, movement direction, speed change and heading change of the target under the first-order straight line model and the second-order curve model respectively specifically includes:

[0017] Substituting the latest position of the target and the time corresponding to the latest position into the first-order straight line model and the second-order curve model respectively, to obtain the smoothed position of the latest position of the target under the first-order straight line model and the smoothed position of the latest position of the target under the second-order curve model;

[0018] Substitute the smoothed position of the latest position of the target under the first-order straight line model into the next cycle prediction time to obtain the predicted position of the target under the first-order straight line model, and substitute the smoothed position of the latest position of the target under the second-order curve model into the next cycle prediction time to obtain the predicted position of the target under the second-order curve model;

[0019] According to the predicted position of the target under the first-order straight line model and the predicted position of the target under the second-order curve model, the movement speed, movement direction, speed change and heading change of the target under the first-order straight line model and the second-order curve model are calculated respectively.

[0020] Furthermore, the final state information includes a final smoothed position, a final speed and a final heading.

[0021] Furthermore, the first-order straight line model and the second-order curve model are given different weights according to the speed change, size and heading change of the target, and weighted average is performed to obtain the final state information of the target, including:

[0022] Establish a three-dimensional coordinate system for speed change, target size and heading change, and set the parameter ranges for speed change, target size and heading change to form a three-dimensional cuboid;

[0023] Define the function f(m,n,p):

[0024]

[0025] Among them, m represents the target size, n represents the heading change, p represents the speed change, and q max and q min are the maximum and minimum values ​​of f(m,n,p), respectively, and q max and q min Pre-set;

[0026] Project the point (m0, n0, p0) of the target to be tracked in the three-dimensional coordinate system onto the diagonal line of the point (m2, n2, p2) and the point (m1, n1, p1) to obtain the point (m3, n3, p3). Substituting it into the formula f(m, n, p) can obtain the corresponding model selection coefficient q = f(m3, n3, p3), where m1 represents the maximum target size, m2 represents the minimum target size, n1 represents the minimum heading change, n2 represents the maximum heading change, p1 represents the minimum speed change, p2 represents the maximum speed change, and q represents the weight of the second-order curve model;

[0027] The fitting result of the first-order straight line model and the fitting result of the second-order curve model are weighted averaged according to the selection coefficient q to obtain the final smooth position, speed and heading of the target to be tracked.

[0028] Furthermore, the q max is 0.5, the q min is 0.

[0029] On the other hand, the present invention also provides a target tracking device based on line segment fitting, the device comprising:

[0030] A line segment fitting module is used to extract several positions at the end of the target's motion trajectory and the corresponding time series, and perform line segment fitting on the target, wherein the line segment fitting includes first-order straight line fitting and second-order curve fitting;

[0031] The motion state acquisition module is used to substitute the current position of the target and the corresponding time according to the fitted first-order straight line model and second-order curve model to calculate the motion speed, motion direction, speed change and heading change of the target under the first-order straight line model and the second-order curve model respectively;

[0032] The target position tracking module is used to assign different weights to the first-order straight line model and the second-order curve model according to the speed change, size and heading change of the target and perform weighted average to obtain the final smooth position, final speed and final heading of the target.

[0033] On the other hand, the present invention also provides a computer device, comprising a processor and a memory, wherein the memory stores a computer program, and the computer program is loaded and executed by the processor to implement any of the above-mentioned target tracking methods based on line segment fitting.

[0034] On the other hand, the present invention further provides a computer-readable storage medium, in which a computer program is stored. The computer program is loaded and executed by a processor to implement any of the above-mentioned target tracking methods based on line segment fitting.

[0035] The beneficial effects of the present invention are:

[0036] (1) The target tracking method based on line segment fitting provided by the present invention has good processing effect on various targets, can output relatively high-quality tracks, and obtain a good compromise in terms of smoothness and fit. The target track generated by this method is consistent with the corresponding ground indicator line, and the single-point error is low.

[0037] (2) The target tracking method based on line segment fitting provided by the present invention can determine the motion state of the target based on information such as target size, acceleration, heading, etc., with high accuracy, and can be used as a basis for setting weighting coefficients in the first-order and second-order models. Moreover, it can also be used as a reference in subsequent target recognition, regional CFAR and other functions.

[0038] (3) The target tracking method based on line segment fitting provided by the present invention can adapt to multiple types of targets, including high-maneuverability targets, high-acceleration targets, etc. For targets moving in a straight line or with low maneuverability, a straight line model is preferred, which has an excellent smoothing effect; for turning targets or high-maneuverability targets, a curve model is preferred, which has low track inertia and strong fit. For targets moving at variable speeds, a short model is preferred to adapt to their acceleration; for targets with stable speeds, a long model is preferred, which has better smoothness.

[0039] (4) The target tracking method based on line segment fitting provided by the present invention has simple logic and strong stability. It can easily and safely perform open-loop intervention in response to various areas and targets in the scene, and perform targeted processing based on a posteriori information. At the same time, in the case of long-term operation, no logical problems such as "runaway, loss of lock" occur. BRIEF DESCRIPTION OF THE DRAWINGS

[0040] Figure 1 is a schematic flow chart of a target tracking method based on line segment fitting provided by an embodiment of the present invention;

[0041] Figure 2 is a schematic diagram of a final state information model established in an embodiment of the present invention;

[0042] Figure 3 It is a two-dimensional radar detection point effect diagram of a small target and a large target provided by an embodiment of the present invention;

[0043] Figure 4 It is a global map of the original points and tracks of small and large targets provided by an embodiment of the present invention;

[0044] Figure 5 This is a diagram showing the effect of a small target maneuvering at a large angle according to an embodiment of the present invention;

[0045] Figure 6 This is a rendering of a large target during a 90-degree turn maneuver according to an embodiment of the present invention;

[0046] Figure 7 This is a diagram showing the effect of a small target in a straight-line motion according to an embodiment of the present invention;

[0047] Figure 8 This is a rendering of the linear motion of a large target according to an embodiment of the present invention;

[0048] Fig. 9 1. It is a schematic diagram comparing the effects of large-scale target turning and three-model Kalman algorithms in an embodiment of the present invention;

[0049] Fig.10 1. It is a schematic diagram comparing the effects of a small target moving straight and three-model Kalman algorithms in an embodiment of the present invention;

[0050] Fig.11 It is a structural block diagram of a target tracking device based on line segment fitting provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0051] The following describes the embodiments of the present invention by specific examples, and those skilled in the art can easily understand other advantages and effects of the present invention from the contents disclosed in this specification. The present invention can also be implemented or applied through other different specific embodiments, and the details in this specification can also be modified or changed in various ways based on different viewpoints and applications without departing from the spirit of the present invention. It should be noted that the following embodiments and features in the embodiments can be combined with each other without conflict.

[0052] Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making any creative work shall fall within the scope of protection of the present invention.

[0053] The problem of target tracking can be divided into three aspects: first, the inherent problems of radar, such as random errors of radar and side lobes generated by algorithms; second, some specific scenarios, such as multipath and occlusion, which lead to reduced or even failure of radar detection efficiency. The above two points will lead to distortion, offset, missed alarm and false alarm of detected target information; third, the motion state of the target itself is diverse, such as stationary, slow, high speed, turning and straight, and the radar cannot detect the target all the time due to the limitation of scanning cycle. In essence, it only collects the target information at a certain moment and cannot fully cope with the sudden change or nonlinear change of the target motion state.

[0054] Since the target not only moves in a straight line, but also maneuvers. To deal with this situation, other algorithms such as Kalman can design three models: left turn, right turn, and straight go. According to the target's motion trajectory, the optimal model is found among the three models and the model is switched. However, the disadvantage is that the selection of parameters for different scenarios and targets lacks universality.

[0055] In order to solve the above technical problems, the following embodiments of the target tracking method, device, equipment and medium based on line segment fitting of the present invention are proposed.

[0056] Example 1

[0057] The problem of target tracking can be divided into three aspects: first, the inherent problems of radar, such as random errors of radar and side lobes generated by algorithms; second, some specific scenarios, such as multipath and occlusion, which lead to reduced or even failure of radar detection efficiency. The above two points will lead to distortion, offset, missed alarm and false alarm of detected target information; third, the motion state of the target itself is diverse, such as stationary, slow, high speed, turning and straight, and the radar cannot detect the target all the time due to the limitation of scanning cycle. In essence, it only collects the target information at a certain moment and cannot fully cope with the sudden change or nonlinear change of the target motion state.

[0058] From the above analysis and actual processing results, if a single state matrix or parameter is used to smooth the track, then the "perfect smoothness" in the ideal state is not obtainable. In the actual process, for targets with different maneuvers, it is impossible to ensure both smoothness and fast response when the target is maneuvering. The more natural way to solve this problem is to use dynamic models and dynamic parameters to adjust adaptively according to target attributes, motion status, etc.

[0059] This embodiment divides the target motion trajectory into two categories: linear motion or curved motion. For the smoothing of linear motion targets, various algorithms essentially perform operations similar to the least squares method to find the line segment closest to each position coordinate during the motion process, that is, the sum of the variances of the line segment and all position points is the smallest.

[0060] Since the target not only moves in a straight line, but also maneuvers. In order to deal with this kind of situation, other algorithms such as Kalman can design three models of left turn, right turn and straight going, find the optimal model among the three types of models according to the target motion trajectory, and switch the models. However, the disadvantage is still the lack of universality in parameter selection for different scenes and different targets. To solve this problem, this embodiment adds a curve model, that is, adds a second-order curve fitting, and finds the second-order curve with the smallest sum of variances with all position points.

[0061] At the same time, when fitting a straight line or curve, the length of the original points used will also affect the fitting effect. The more and longer the original points are, the greater the stability and the lower the flexibility.

[0062] The first-order straight line, second-order curve and fitting of different lengths can all express the local optimal target trajectory. We need to select the fitting method and length to obtain the most appropriate target tracking trajectory.

[0063] Theoretically, a higher-dimensional third-order curve has a better effect on the target of "S"-shaped motion trajectory. But in fact, a second-order curve with appropriate parameters can meet most needs. Higher-order curve fitting can achieve higher fit, but the smoothness will inevitably decrease.

[0064] Reference Figure 1 ,like Figure 1 FIG. 1 is a flow chart of a target tracking method based on line segment fitting provided in this embodiment, and the method specifically comprises the following steps:

[0065] Step S100: extracting several positions at the end of the target's motion trajectory and the corresponding time series, and performing line segment fitting on the target. The line segment fitting includes first-order straight line fitting and second-order curve fitting.

[0066] Specifically, first, extract the positions (x0, y0), (x1, y1), (x2, y2)...(x n ,y n ), and its corresponding time series t, the first-order straight line and second-order curve fitting are performed on the X-axis and Y-axis respectively, and the fitting results are as follows:

[0067] First-order straight line fit:

[0068] x L =b1*t+c1;

[0069] y L =b2*t+c2;

[0070] Second-order curve fitting:

[0071] x C =a3*t 2 +b3*t+c3;

[0072] y C =a4*t 2 +b4*t+c4;

[0073] Step S200: Substitute the current position of the target and the corresponding time into the fitted first-order straight line model and second-order curve model to calculate the target's movement speed, movement direction, speed change and heading change under the first-order straight line model and the second-order curve model respectively.

[0074] This step specifically includes:

[0075] Substitute the latest position of the target and the time corresponding to the latest position into the first-order straight line model and the second-order curve model respectively to obtain the smoothed position of the latest position of the target under the first-order straight line model and the smoothed position of the latest position of the target under the second-order curve model.

[0076] Substitute the smoothed position of the target's latest position under the first-order straight line model into the next cycle prediction time to obtain the target's predicted position under the first-order straight line model, and substitute the smoothed position of the target's latest position under the second-order curve model into the next cycle prediction time to obtain the target's predicted position under the second-order curve model.

[0077] According to the predicted position of the target under the first-order straight line model and the predicted position of the target under the second-order curve model, the target's movement speed, movement direction, speed change and heading change under the first-order straight line model and the second-order curve model are calculated respectively.

[0078] Taking the curve model as an example, substitute the latest target position (x n ,y n ) corresponds to the time t n , we can get the smooth position of the curve model at this point

[0079] Substitute the next cycle prediction time t n+1 =t n +T, T is the radar scanning period, and the predicted position of the target under the curve model can be obtained

[0080] In curve mode, the target movement speed and direction can be calculated:

[0081]

[0082]

[0083] Based on this, the change in target movement speed and target movement heading can be calculated:

[0084]

[0085]

[0086] The above completes the n Smooth position, speed, direction, speed change, heading change and t n+1 Predict the position at any moment. Similarly, the corresponding parameter motion speed under the first-order linear model can be obtained Direction of movement Speed ​​change Course Change

[0087] Step S300: assign different weights to the first-order straight line model and the second-order curve model according to the speed change, size and heading change of the target and perform weighted averaging to obtain the final state information of the target.

[0088] Specifically, in order to determine the motion state of the target, this embodiment makes the following settings based on objective physical properties and motion laws:

[0089] Targets that are accelerating or decelerating tend to move in a straight line rather than making large turns.

[0090] The larger the target, the greater its inertia and the worse its maneuverability.

[0091] According to the above settings, this embodiment defines: a large-sized target is in a motion state with small speed changes and small heading changes, which is a low-maneuverability motion state, and is more inclined to perform a more stable straight-line motion; a small-sized target is in a motion state with large speed changes and large heading changes, which is a high-maneuverability motion state, and is more inclined to perform a more flexible curved motion.

[0092] Reference Figure 2 ,like Figure 2 The figure shows a schematic diagram of the final state information model established in this embodiment. As shown in the figure above, the range of parameters such as speed change, size, and heading change are set to form a three-dimensional cuboid, in which:

[0093] The x-axis is the target size m, in meters, which is the square root of the sum of the target's distance dimension length and angle dimension width.

[0094] The y-axis is the heading change n, in degrees / second, which describes the variability of the target's motion direction.

[0095] The z-axis is the speed change p, in meters per second 2 .

[0096] The upper left corner point (m2,n2,p2) has the highest maneuverability, and the lower right corner point (m1,n1,p1) has the lowest maneuverability. The parameter settings can be determined according to the physical properties of the target or the actual project. In this figure, the target size m∈[10,30], the heading change n∈[2,15], and the speed change p∈[1,10] are set. If it exceeds this range, it is set to the boundary value. Therefore, (m2,n2,p2)=(10,15,10), (m1,n1,p1)=(30,2,1)

[0097] Different weights are assigned to the first-order linear model and the second-order curve model according to the speed change, size and heading change of the target, and weighted average is performed to obtain the final state information of the target, including:

[0098] Establish a three-dimensional coordinate system for speed change, target size and heading change, and set the parameter ranges for speed change, target size and heading change to form a three-dimensional cuboid;

[0099] Define the function f(m,n,p):

[0100]

[0101] Among them, m represents the target size, n represents the heading change, p represents the speed change, and q max and q min are the maximum and minimum values ​​of f(m,n,p), respectively, and q max and q min Preset.

[0102] Project the point (m0, n0, p0) of the target to be tracked in the three-dimensional coordinate system onto the diagonal line of the point (m2, n2, p2) and the point (m1, n1, p1) to obtain the point (m3, n3, p3). Substituting it into the formula f(m, n, p) can obtain the corresponding model selection coefficient q = f(m3, n3, p3), where m1 represents the maximum target size, m2 represents the minimum target size, n1 represents the minimum heading change, n2 represents the maximum heading change, p1 represents the minimum speed change, p2 represents the maximum speed change, and q represents the weight of the second-order curve model.

[0103] The fitting results of the first-order straight line model and the second-order curve model are weighted averaged according to the selection coefficient q to obtain the final smooth position, speed and heading of the target to be tracked.

[0104] For any mobility state, that is, any point (m0, n0, p0) in the cuboid, project it onto the diagonal line where the points (m2, n2, p2) and (m1, n1, p1) are located, and the resulting point (m3, n3, p3) is substituted into the formula f(m, n, p) to obtain the corresponding model selection coefficient q = f(m3, n3, p3).

[0105] Therefore, according to the model selection coefficient q, the weighted sum of the first-order straight line and second-order curve models can be obtained. n The final smooth position (x n ,y n ), speed V n , heading D n :

[0106]

[0107]

[0108]

[0109]

[0110] Similarly, q max ,q min Set as the longest and shortest length of the model fitting length, and also obtain the length of the number of original points used in fitting.

[0111] As an implementation method, in this embodiment, the model selection coefficient q can be set add ∈[0.0,0.5], that is, setting q max is 0.5, q min 0. It means that when the target is in the lowest mobility state, the model selection coefficient is set to 0.0, that is, the first-order straight line model is fully trusted; when the target is in the highest mobility state, the model selection coefficient is set to 0.5, and the fitting results of the first-order straight line model and the second-order curve model are weighted averaged by the coefficient 0.5.

[0112] The target tracking method based on line segment fitting provided in this embodiment has good processing effect on various types of targets, can output high-quality tracks, and obtain a good compromise in smoothness and fit. The target track generated by this method is consistent with the corresponding ground indicator line, and the single-point error is low. This method can determine the motion state of the target based on information such as target size, acceleration, heading, etc., with high accuracy, and can be used as a basis for setting weighting coefficients in first-order and second-order models. Moreover, it can also be used as a reference for other functions such as target identification and regional CFAR in the future. This method can adapt to multiple types of targets, including high-maneuverability targets, high-acceleration targets, etc. For targets with linear motion or low-maneuverability targets, a straight line model tends to be used, and the smoothing effect is excellent; for turning targets or high-maneuverability targets, a curve model tends to be used, and the track inertia is low and the fit is strong. For targets with variable speed motion, a short model tends to be used to adapt to its acceleration; for targets with stable speed, a long model tends to be used, and the smoothness is better. This method has simple logic and strong stability. In response to various areas and targets in the scene, open-loop intervention can be carried out simply and safely, and targeted processing can be carried out based on a posteriori information. At the same time, in the case of long-term operation, no logical problems such as "runaway, loss of lock" have occurred.

[0113] Example 2

[0114] Reference Figure 3 ,like Figure 3 The following is a two-dimensional radar detection point effect diagram of a small target and a large target provided by this embodiment. The diagram shows the radar detection original points in a certain scanning cycle and the condensation center after the condensation algorithm. The upper left corner is a small target (vehicle), and the lower right corner is a large target (aircraft). The "." point is the original detection point, and the "o" point is the condensation center. The original point position used in the tracking stage of this embodiment is the condensation center, and the size used is obtained according to the distribution range of the original detection points.

[0115] Reference Figure 4 ,like Figure 4Shown is the global graph of the original point and track output of the small target and large target provided by this embodiment. In this embodiment, the global original point center position and tracking effect diagram of a small target and a large target. In this embodiment, the radar software system replays the measured data, saves the data before processing and the results after processing, and analyzes and draws the graph by matlab software. The radar data beat is 1 second. Among them, there is a loss phenomenon caused by occlusion during the whole operation of the small target, and there are motion states such as straight running, high maneuverability and stationary stop. During the whole operation of the large target, there are motion states such as rapid deceleration, straight running, turning, etc. The information such as the original point size of the target is shown in Table 1.

[0116]

[0117] Table 1 Original point size information table of the target

[0118] The parameters set, the model fitting length is q len ∈[6,12], that is, the fitting length range is 6 to 12 beats. The model selection coefficient is qadd∈[0,0.5], that is, when the mobility is the weakest, the first-order straight line model is fully trusted, and when the mobility is the strongest, the first-order straight line and the second-order curve are used with average weighting. During the entire target operation, the statistical parameters are shown in Table 2:

[0119]

[0120]

[0121] Table 2 Target operation parameter statistics

[0122] Except for the take-off track stage, large targets tend to use a more stable long straight line model in other states; small targets have strong maneuverability, the fitting length used in the whole process is relatively short, and the model selection coefficient is relatively high. Especially when turning, they tend to use a second-order curve model to reduce "outward impact" and enhance track fit.

[0123] Reference Figure 5 ,like Figure 5 The following is a diagram showing the effect of a small target in a large-angle maneuver according to this embodiment, and its parameter information is shown in Table 3:

[0124]

[0125] Table 3 Large-angle maneuvering parameter information of small targets

[0126] Beat 9 to beat 29, so the target completes a large-angle maneuver in 21 seconds. The model selection coefficient has two "first increase and then decrease" processes, and the model fitting length changes accordingly. The changes of the two are consistent with the actual maneuver trajectory of the target. The difference between the position of the track point and the original point after processing is less than 5 meters, the inertia is low, the "outward impact" is weak, and the track effect is smooth. After high maneuvering, the target stays still at beats 29 to 91, and moves again at beat 92. It can be seen that this method has excellent operation effect in various motion states.

[0127] Reference Figure 6 ,like Figure 6 The figure shows the effect of a 90-degree turning maneuver of a large target in this embodiment, and its parameter information is shown in Table 4:

[0128]

[0129] Table 4 90 degree maneuver parameter information table of large targets

[0130] From beat 84 to beat 107, the target completed a 90-degree turn in 25 seconds. The target turned smoothly, but the radar detection effect was affected due to the change in the target's attitude, and the target position jittered slightly. The model selection coefficient was less than 0.1, and the straight line model was trusted, and the track effect was smooth and stable.

[0131] Reference Figure 7 ,like Figure 7 The following is a diagram showing the effect of the small target in the linear motion of this embodiment, and its parameter information is shown in Table 5:

[0132]

[0133] Table 5 Linear motion parameter information table of small targets

[0134] Due to the small target size, the model selection coefficient is not reduced to 0.0, but it is still at a low level, which effectively reduces the jitter.

[0135] Reference Figure 8 ,like Figure 8 The figure shows the effect of the linear motion of the large target in this embodiment, and its parameter information is shown in Table 6:

[0136]

[0137] Table 6 Linear motion parameter information table of large targets

[0138] It can be seen from the above table that the model selection coefficient is the lowest value of 0.0 for a long time. The straight line model is fully trusted and the track smoothing effect is excellent.

[0139] Reference Fig. 9 ,like Fig. 9The figure shows a schematic diagram of the comparison between the large target turning and the three-model Kalman algorithm in this embodiment. The comparison data uses the turning process of the large target track with a beat of 75 to 115. The ground indicator line of the scene map corresponding to the beat is used as the standard track. It is assumed that the target moves along the ground indicator line. The standard deviation of the distance difference between the beat data corresponding to each algorithm and the ground indicator line is calculated. The comparison model is the three-model Kalman, which includes three states: straight line, left turn, and right turn, and can be switched adaptively. The algorithm platform is matlab, and two parameter settings of low smoothness and high smoothness are set in the comparison. Figure 7 The lower right figure shows the model coefficients of the three-model Kalman algorithm. It can be seen that it has determined the left-turn trend at beat 80. In theory, the algorithm is better than the standard single-model Kalman. However, the high-smoothness Kalman still has strong inertia and is prone to strong "outward impact" after the turn is completed. The smoothing effect of the low-smoothness Kalman is average. This method has significantly better smoothness while ensuring fit. In addition, during the comparison process, the Kalman occasional iteration matrix singularity leads to iteration errors.

[0140] Reference Fig.10 ,like Fig.10 The figure shows the comparison between the small target straight movement and the three-model Kalman algorithm in this embodiment. The comparison data uses the straight movement process of the small target with a track beat of 100 to 140, and the comparison example is the three-model Kalman. This method is obviously better in fit and smoothness.

[0141] Example 3

[0142] Reference Fig.11 ,like Fig.11 FIG. 1 is a block diagram of a target tracking device based on line segment fitting provided in this embodiment, and the device includes:

[0143] A line segment fitting module 10 is used to extract several positions at the end of the target's motion trajectory and the corresponding time series, and perform line segment fitting on the target, wherein the line segment fitting includes first-order straight line fitting and second-order curve fitting;

[0144] The motion state acquisition module 20 is used to calculate the motion speed, motion direction, speed change and heading change of the target under the first-order straight line model and the second-order curve model respectively by substituting the current position of the target and the corresponding time according to the fitted first-order straight line model and the second-order curve model;

[0145] The target position tracking module 30 is used to assign different weights to the first-order straight line model and the second-order curve model according to the speed change, size and heading change of the target and perform weighted averaging to obtain the final smooth position, final speed and final heading of the target.

[0146] The beneficial effects of the target tracking device based on line segment fitting provided by this embodiment are detailed in the previous embodiments and will not be described again here.

[0147] Example 4

[0148] This preferred embodiment provides a computer device, which can implement the steps in any embodiment of the target tracking method based on line segment fitting provided in the embodiments of the present application. Therefore, the beneficial effects of the target tracking method based on line segment fitting provided in the embodiments of the present application can be achieved. Please refer to the previous embodiments for details and will not be repeated here.

[0149] Example 5

[0150] Those skilled in the art can understand that all or part of the steps in the various methods of the above embodiments can be completed by instructions, or by controlling related hardware through instructions, and the instructions can be stored in a computer-readable storage medium and loaded and executed by a processor. To this end, an embodiment of the present invention provides a storage medium, which stores a plurality of instructions, and the instructions can be loaded by a processor to execute the steps of any embodiment of the target tracking method based on line segment fitting provided by the embodiment of the present invention.

[0151] The storage medium may include: a read-only memory (ROM), a random access memory (RAM), a magnetic disk or an optical disk, etc.

[0152] Since the instructions stored in the storage medium can execute the steps in any of the target tracking method embodiments based on line segment fitting provided in the embodiments of the present invention, the beneficial effects that can be achieved by any of the target tracking method embodiments based on line segment fitting provided in the embodiments of the present invention can be achieved. For details, please refer to the previous embodiments and will not be repeated here.

[0153] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions and improvements made within the spirit and principles of the present invention should be included in the protection scope of the present invention.

Claims

1. A target tracking method based on line segment fitting, characterized in that: The method comprises: Extracting several positions at the end of the target's motion trajectory and corresponding time series, and performing line segment fitting on the target, wherein the line segment fitting includes first-order straight line fitting and second-order curve fitting; Substitute the current position of the target and the corresponding time into the fitted first-order straight line model and second-order curve model to calculate the target's movement speed, movement direction, speed change and heading change under the first-order straight line model and second-order curve model respectively; Different weights are assigned to the first-order straight line model and the second-order curve model according to the speed change, size and heading change of the target, and weighted average is performed to obtain the final state information of the target, including: Establish a three-dimensional coordinate system for speed change, target size and heading change, and set the parameter ranges for speed change, target size and heading change to form a three-dimensional cuboid; Define the function f(m,n,p): Among them, m represents the target size, n represents the heading change, p represents the speed change, and q max and q min are the maximum and minimum values ​​of f(m,n,p), respectively, and q max and q min Pre-set; Project the point (m0, n0, p0) of the target to be tracked in the three-dimensional coordinate system onto the diagonal line of the point (m2, n2, p2) and the point (m1, n1, p1) to obtain the point (m3, n3, p3). Substituting it into the formula f(m, n, p) can obtain the corresponding model selection coefficient q = f(m3, n3, p3), where m1 represents the maximum target size, m2 represents the minimum target size, n1 represents the minimum heading change, n2 represents the maximum heading change, p1 represents the minimum speed change, p2 represents the maximum speed change, and q represents the weight of the second-order curve model; The fitting result of the first-order straight line model and the fitting result of the second-order curve model are weighted averaged according to the selection coefficient q to obtain the final smooth position, speed and heading of the target to be tracked.

2. The target tracking method based on line segment fitting according to claim 1, characterized in that: The first-order straight line and second-order curve fitting of the target specifically includes: Extracting a plurality of positions at the end of the target motion trajectory and corresponding time series of the plurality of positions; The first-order straight line fitting and the second-order curve fitting are performed on the X-axis and the Y-axis respectively.

3. The target tracking method based on line segment fitting according to claim 1, characterized in that: Substituting the fitted first-order straight line model and second-order curve model into the current position of the target and the corresponding time to calculate the movement speed, movement direction, speed change and heading change of the target under the first-order straight line model and the second-order curve model respectively specifically includes: Substituting the latest position of the target and the time corresponding to the latest position into the first-order straight line model and the second-order curve model respectively, to obtain the smoothed position of the latest position of the target under the first-order straight line model and the smoothed position of the latest position of the target under the second-order curve model; Substitute the smoothed position of the latest position of the target under the first-order straight line model into the next cycle prediction time to obtain the predicted position of the target under the first-order straight line model, and substitute the smoothed position of the latest position of the target under the second-order curve model into the next cycle prediction time to obtain the predicted position of the target under the second-order curve model; According to the predicted position of the target under the first-order straight line model and the predicted position of the target under the second-order curve model, the movement speed, movement direction, speed change and heading change of the target under the first-order straight line model and the second-order curve model are calculated respectively.

4. The target tracking method based on line segment fitting according to claim 1, characterized in that: The final state information includes a final smoothed position, a final speed, and a final heading.

5. The target tracking method based on line segment fitting according to claim 1, characterized in that: The q max is 0.5, the q min is 0.

6. A target tracking device based on line segment fitting, characterized in that: The device comprises: A line segment fitting module is used to extract several positions at the end of the target's motion trajectory and the corresponding time series, and perform line segment fitting on the target, wherein the line segment fitting includes first-order straight line fitting and second-order curve fitting; The motion state acquisition module is used to substitute the current position of the target and the corresponding time according to the fitted first-order straight line model and second-order curve model to calculate the motion speed, motion direction, speed change and heading change of the target under the first-order straight line model and the second-order curve model respectively; The target position tracking module is used to assign different weights to the first-order straight line model and the second-order curve model according to the speed change, size and heading change of the target and perform weighted average to obtain the final smooth position, final speed and final heading of the target, including: Establish a three-dimensional coordinate system for speed change, target size and heading change, and set the parameter ranges for speed change, target size and heading change to form a three-dimensional cuboid; Define the function f(m,n,p): Among them, m represents the target size, n represents the heading change, p represents the speed change, and q max and q min are the maximum and minimum values ​​of f(m,n,p), respectively, and q max and q min Pre-set; Project the point (m0, n0, p0) of the target to be tracked in the three-dimensional coordinate system onto the diagonal line of the point (m2, n2, p2) and the point (m1, n1, p1) to obtain the point (m3, n3, p3). Substituting it into the formula f(m, n, p) can obtain the corresponding model selection coefficient q = f(m3, n3, p3), where m1 represents the maximum target size, m2 represents the minimum target size, n1 represents the minimum heading change, n2 represents the maximum heading change, p1 represents the minimum speed change, p2 represents the maximum speed change, and q represents the weight of the second-order curve model; The fitting result of the first-order straight line model and the fitting result of the second-order curve model are weighted averaged according to the selection coefficient q to obtain the final smooth position, speed and heading of the target to be tracked.

7. A computer device, characterized in that: The computer device includes a processor and a memory, wherein a computer program is stored in the memory, and the computer program is loaded and executed by the processor to implement the target tracking method based on line segment fitting as described in any one of claims 1 to 5.

8. A computer-readable storage medium, characterized in that: The storage medium stores a computer program, which is loaded and executed by a processor to implement the target tracking method based on line segment fitting as described in any one of claims 1 to 5.

Images