CN-121350399-B - Pre-detection tracking method and related products

CN121350399BCN 121350399 BCN121350399 BCN 121350399BCN-121350399-B

Abstract

The application discloses a tracking method before detection and a related product, wherein the method comprises the steps of determining a tracking algorithm before detection and a state result of a tracking target at the last moment; the method comprises the steps of obtaining a state result of a current moment based on a state result of a tracking algorithm before detection and a state result of a tracking target at a last moment, constructing a correlation gate of the current moment based on the state result of the last moment when the state result of the last moment meets a preset condition, obtaining a tracking result of the tracking target based on the state result of the tracking algorithm before detection, the state result of the current moment and the correlation gate of the current moment, and screening at least one of the state result of the current moment and a measurement data set of the current moment by the correlation gate of the current moment. The method fundamentally avoids the problems of weight offset and track drift caused by the fact that interference signals such as strong clutter, outlier points and the like are taken into account simultaneously in the traditional TBD method, and improves the reliability of the tracking method before detection.

Inventors

PAN MENGGUAN
XUE ANQI
ZHOU XIAO
ZHAO LUYU
LI YINGSONG
HUANG ZHIXIANG
LIU PENG
QI WANGDONG

Assignees

安徽大学

Dates

Publication Date: 20260508
Application Date: 20251218

Claims (9)

1. A pre-detection tracking method, characterized in that the pre-detection tracking method comprises: Determining a tracking algorithm before detection and a state result of a tracking target at the last moment; based on the tracking algorithm before detection and the state result of the tracking target at the last moment, obtaining the state result at the current moment; under the condition that the state result of the last moment meets the preset condition, constructing an association gate of the current moment based on the state result of the last moment; The method comprises the steps of obtaining a tracking result of a tracking target based on a tracking algorithm before detection, a state result at the current moment and an association gate at the current moment, wherein the association gate at the current moment is used for screening at least one of the state result at the current moment and a measurement data set at the current moment; The obtaining the tracking result of the tracking target based on the tracking algorithm before detection, the state result of the current moment and the association gate of the current moment specifically includes: acquiring a measurement data set of the tracking target at the current moment, wherein the measurement data set comprises a plurality of measurement points and measurement data of each measurement point; Reserving measuring points in the range of the association gate of the current moment in the measuring data set of the current moment, deleting measuring points not in the range of the association gate of the current moment in the measuring data set of the current moment, and obtaining a screened measuring data set; If the pre-detection tracking algorithm is a pre-detection tracking algorithm based on particle filtering, a tracking result of the tracking target is obtained based on a weighted average method, the state result at the current moment and the screened measurement data set; If the pre-detection tracking algorithm is a pre-detection tracking algorithm based on dynamic programming, a tracking result of the tracking target is obtained based on the state result of the current moment, the association gate of the current moment and the screened measurement data set; and if the pre-detection tracking algorithm is a pre-detection tracking algorithm based on the Hough transformation, obtaining a tracking result of the tracking target based on the Hough transformation and the screened measurement data set.
2. The pre-detection tracking method according to claim 1, wherein the obtaining the state result at the current time based on the pre-detection tracking algorithm and the state result at the previous time of the tracking target specifically includes: if the pre-detection tracking algorithm is a pre-detection tracking algorithm based on particle filtering, a state result of the tracking target at the last moment is characterized as a particle set result of the tracking target at the last moment, wherein the particle set result comprises posterior state estimation and weight of a plurality of particles; The posterior state estimation of each particle in the particle set result at the previous moment of the tracking target is transmitted through a motion model, noise is superimposed in the transmission process, and the prior state estimation of each particle at the current moment is obtained; Based on prior state estimation of a plurality of particles at the current moment, a state result at the current moment is obtained.
3. The pre-detection tracking method according to claim 1, wherein the obtaining the state result at the current time based on the pre-detection tracking algorithm and the state result at the previous time of the tracking target specifically includes: If the pre-detection tracking algorithm is a pre-detection tracking algorithm based on dynamic programming, the state result of the last moment of the tracking target is represented as a scoring matrix and a pointer matrix of the last moment of the tracking target, wherein the scoring matrix comprises scoring results of a plurality of discrete states; the state transition probability of each discrete state at the previous moment is respectively input into a motion model, and the transition score of each discrete state output by the motion model and each discrete state at the current moment are respectively obtained; updating the accumulated score of the corresponding discrete state by using the transfer score to obtain the accumulated score of each discrete state at the current moment; screening the discrete states at the current moment by using the accumulated score threshold value to obtain the discrete states screened at the current moment; updating the scoring matrix and the pointer matrix at the previous moment based on the plurality of discrete states screened at the current moment to obtain the scoring matrix and the pointer matrix at the current moment; and taking the scoring matrix and the pointer matrix at the current moment as a state result at the current moment.
4. The pre-detection tracking method according to claim 1, wherein the obtaining the state result at the current time based on the pre-detection tracking algorithm and the state result at the previous time of the tracking target specifically includes: if the pre-detection tracking algorithm is a pre-detection tracking algorithm based on Hough transform, the state result of the last moment of the tracking target is represented as the detection result of the last moment of the tracking target, wherein the detection result comprises a plurality of track parameters, and the track parameters comprise coordinates and speed; determining an optimal track parameter in a plurality of track parameters at the previous moment; predicting track parameters at the current moment based on the optimal track parameters; and adding the track parameter at the current moment into the detection result at the previous moment to obtain a state result at the current moment.
5. The pre-detection tracking method according to claim 1, wherein when the state result of the previous time meets a preset condition, constructing an association gate at the current time based on the state result of the previous time specifically includes: If the state result of the previous moment meets the preset condition, extracting the state result of the previous moment to obtain position data of the previous moment, wherein the position data comprises position components or coordinates; calculating the statistical characteristics of the state result at the last moment on the spatial distribution; Obtaining the size of the associated door according to the statistical characteristics and a preset confidence coefficient; and taking the position data of the previous moment as a center, taking the associated gate size as a range to divide geometric areas, and taking the geometric areas obtained by division as the associated gate of the current moment.
6. The pre-detection tracking method according to claim 1, wherein if the pre-detection tracking algorithm is a pre-detection tracking algorithm based on particle filtering, the tracking result of the tracking target is obtained based on a weighted average method, the state result at the current time and the filtered measurement dataset, and specifically comprises: If the pre-detection tracking algorithm is a pre-detection tracking algorithm based on particle filtering, updating the weight of each particle in the state result at the current moment by using the screened measurement data set to obtain a state result updated at the current moment; And obtaining a tracking result of the tracking target based on the updated state result at the current moment by a weighted average method, wherein the tracking result comprises posterior state estimation at the current moment and posterior existence probability at the current moment.
7. The pre-detection tracking method according to claim 1, wherein if the pre-detection tracking algorithm is a pre-detection tracking algorithm based on dynamic programming, the tracking result of the tracking target is obtained based on the state result of the current time, the correlation gate of the current time and the screened measurement dataset, and specifically includes: if the pre-detection tracking algorithm is based on dynamic programming, screening the state result at the current moment by using the association gate at the current moment to obtain the state result screened at the current moment; Scoring each discrete state in the state result screened at the current moment based on the screened measurement data set to obtain a final state result; Optimizing the final state result to obtain an optimal discrete state; taking the optimal discrete state as an end point, and backtracking each discrete state in the final state result based on a reference matrix in the state result at the current moment to obtain an optimal track; and taking the optimal track as a tracking result of the tracking target.
8. The pre-detection tracking method according to claim 1, wherein if the pre-detection tracking algorithm is a pre-detection tracking algorithm based on hough transform, the tracking result of the tracking target is obtained based on hough transform and the screened measurement dataset, and specifically comprises: if the pre-detection tracking algorithm is a pre-detection tracking algorithm based on Hough transform, carrying out Hough transform on each measuring point in the screened measuring data set to obtain a final detection result; optimizing the final detection result to obtain a final track parameter; and taking the coordinates in the final track parameters as a tracking result of the tracking target.
9. A pre-detection tracking device, the pre-detection tracking device comprising: the algorithm determining module is used for determining a tracking algorithm before detection and a state result of a tracking target at the last moment; The prediction module is used for obtaining a state result at the current moment based on the tracking algorithm before detection and a state result at the last moment of the tracking target; The judging module is used for constructing an association gate at the current moment based on the state result at the last moment when the state result at the last moment meets the preset condition; The tracking result determining module is used for obtaining a tracking result of the tracking target based on the tracking algorithm before detection, the state result of the current moment and the association gate of the current moment, wherein the association gate of the current moment is used for screening at least one of the state result of the current moment and the measurement data set of the current moment; The tracking result determining module specifically comprises: The acquisition unit is used for acquiring a measurement data set of the tracking target at the current moment, wherein the measurement data set comprises a plurality of measurement points and measurement data of each measurement point; The screening unit is used for reserving measuring points in the range of the association gate of the current moment in the measuring data set of the current moment, deleting measuring points not in the range of the association gate of the current moment in the measuring data set of the current moment, and obtaining a screened measuring data set; The particle filtering result determining unit is used for obtaining a tracking result of the tracking target based on a weighted average method, the state result of the current moment and the screened measurement data set if the pre-detection tracking algorithm is a pre-detection tracking algorithm based on particle filtering; The dynamic programming result determining unit is used for obtaining a tracking result of the tracking target based on the state result of the current moment, the association gate of the current moment and the screened measurement data set if the pre-detection tracking algorithm is a pre-detection tracking algorithm based on dynamic programming; and the Hough transformation result determining unit is used for obtaining a tracking result of the tracking target based on the Hough transformation and the screened measurement data set if the pre-detection tracking algorithm is a pre-detection tracking algorithm based on the Hough transformation.

Description

Pre-detection tracking method and related products Technical Field The application relates to the technical field of computers, in particular to a tracking method before detection and related products. Background Current pre-detection tracking algorithms (TBD) generally incorporate the signal intensities of all pixels into the calculation at the same time when processing the raw data. This results in a fatal defect that it is impossible to distinguish between signals originating from real targets and measured wild values originating from strong interference. When a outlier (clutter) with intensity far higher than that of the real target appears, the energy accumulation or weight calculation process of the algorithm will be dominated by the outlier, resulting in the final output track being incorrectly positioned to the outlier position, resulting in the loss of the real target. This problem is very common in practical applications (e.g. in the presence of strong static reflectors or electronic interference) and severely limits the reliability of TBD technology. Disclosure of Invention Based on the above problems, the application provides a pre-detection tracking method and related products, and aims to improve the reliability of the pre-detection tracking method. The embodiment of the application discloses the following technical scheme: the first aspect of the present application provides a pre-detection tracking method, which includes: Determining a tracking algorithm before detection and a state result of a tracking target at the last moment; based on the tracking algorithm before detection and the state result of the tracking target at the last moment, obtaining the state result at the current moment; under the condition that the state result of the last moment meets the preset condition, constructing an association gate of the current moment based on the state result of the last moment; and obtaining a tracking result of the tracking target based on the tracking algorithm before detection, the state result at the current moment and the association gate at the current moment, wherein the association gate at the current moment is used for screening at least one of the state result at the current moment and the measurement data set at the current moment. A second aspect of the present application provides a pre-detection tracking apparatus including: the algorithm determining module is used for determining a tracking algorithm before detection and a state result of a tracking target at the last moment; The prediction module is used for obtaining a state result at the current moment based on the tracking algorithm before detection and a state result at the last moment of the tracking target; The judging module is used for constructing an association gate at the current moment based on the state result at the last moment when the state result at the last moment meets the preset condition; the tracking result determining module is used for obtaining a tracking result of the tracking target based on the pre-detection tracking algorithm, the state result of the current moment and the association gate of the current moment, and the association gate of the current moment is used for screening at least one of the state result of the current moment and the measurement data set of the current moment. Compared with the prior art, the application has the following beneficial effects: The method comprises the steps of determining a state result of a tracking algorithm before detection and a state result of a tracking target at the last moment, obtaining a state result of a current moment based on the state result of the tracking algorithm before detection and the state result of the tracking target at the last moment, constructing a correlation gate of the current moment based on the state result of the last moment under the condition that the state result of the last moment meets preset conditions, obtaining the tracking result of the tracking target based on the state result of the tracking algorithm before detection, the state result of the current moment and the correlation gate of the current moment, and screening at least one of the state result of the current moment and the measurement data set of the current moment by the correlation gate of the current moment. The application determines the TBD algorithm and the state result of the tracking target at the last moment, and predicts and generates the state result at the current moment on the basis. When the state result of the last moment meets the preset condition, the full-view field measurement data is not subjected to indiscriminate processing any more, and the associated gate of the current moment is dynamically generated based on the state result of the last moment. The association gate is used for screening the state result, the original measurement data or the combination of the state result and the original measurement data at the current moment, and on