阅读说明：本技术 基于追踪微多普勒图的行人识别方法和装置 (Pedestrian identification method and device based on tracking micro-Doppler image ) 是由 岳靓 陶烨 纪永飞 屈操 于 2021-09-18 设计创作，主要内容包括：本发明提供一种基于追踪微多普勒图的行人识别方法,包括以下步骤：获取雷达信号数据；对所述雷达信号数据进行处理得到雷达点云数据；对所述雷达点云数据进行聚类和目标追踪,得到目标轨迹信息；基于目标轨迹信息筛选得到稳定存在的目标,再对稳定存在的目标进行微多普勒数据的提取,累积多帧有效的微多普勒数据得到目标微多普勒图；选取目标轨迹信息中的目标速度、距离、强度以及累积得到的目标微多普勒图作为有效特征；根据不同类别目标的多组雷达信号数据获得的有效特征训练分类器；利用分类器将各种类别目标的特征分类,筛选出具有行人属性的目标。本发明能够消除环境干扰、提高识别准确度,具有高安全性、多目标适用的特点。(The invention provides a pedestrian identification method based on a tracking micro Doppler image, which comprises the following steps of: acquiring radar signal data; processing the radar signal data to obtain radar point cloud data; clustering and target tracking are carried out on the radar point cloud data to obtain target track information; screening based on target track information to obtain a stably existing target, extracting micro Doppler data of the stably existing target, and accumulating multiple frames of effective micro Doppler data to obtain a target micro Doppler image; selecting target speed, distance, strength and an accumulated target micro-Doppler image in target track information as effective characteristics; training a classifier according to effective characteristics obtained by a plurality of groups of radar signal data of different classes of targets; and classifying the characteristics of various targets by using a classifier, and screening out the targets with pedestrian attributes. The invention can eliminate environmental interference and improve the identification accuracy, and has the characteristics of high safety and multi-target applicability.)

1. A pedestrian identification method based on a tracking micro Doppler image is characterized by comprising the following steps:

acquiring radar signal data;

processing the radar signal data to obtain radar point cloud data;

clustering and target tracking are carried out on the radar point cloud data to obtain target track information;

screening based on target track information to obtain a stably existing target, extracting micro Doppler data of the stably existing target, and accumulating multiple frames of effective micro Doppler data to obtain a target micro Doppler image;

selecting target speed, distance, strength and an accumulated target micro-Doppler image in target track information as effective characteristics;

training a classifier according to effective characteristics obtained by a plurality of groups of radar signal data of different classes of targets;

and classifying the characteristics of various targets by using a classifier, and screening out the targets with pedestrian attributes.

2. The method of claim 1 wherein the method of pedestrian identification based on tracking micro-Doppler maps,

the radar signal data is radar intermediate frequency signal data obtained by processing radar echo signals after reflected radar detection signals are transmitted.

3. The method of claim 1 wherein the method of pedestrian identification based on tracking micro-Doppler maps,

the processing of the radar signal data to obtain radar point cloud data specifically comprises: and performing one-dimensional Fourier transform, two-dimensional Fourier transform, constant false alarm detection and digital beam forming angle measurement signal processing on the radar signal data to obtain radar point cloud data, wherein the radar point cloud data comprises target speed, distance, angle and intensity information.

4. The method of claim 1 wherein the method of pedestrian identification based on tracking micro-Doppler maps,

the screening of the target trajectory information to obtain the stably existing target specifically includes: and screening to obtain the stably existing target according to the existing length of the target track.

5. The method of claim 1 wherein the method of pedestrian identification based on tracking micro-Doppler maps,

the extracting of the micro-doppler data of the stably existing target is performed, and the multi-frame effective micro-doppler data are accumulated to obtain a target micro-doppler image, which specifically comprises the following steps:

initializing an accumulation matrix, wherein the accumulation matrix is used for storing accumulated micro Doppler data and target track information;

matching the tracked target track information with the target track information accumulated in the accumulation matrix, wherein the target track information specifically comprises inner circulation operation and outer circulation operation;

in the inner loop operation, matching the target track information obtained by tracking the ith target with the jth target track information in the accumulation matrix; the matching results are divided into three categories: the first type of matching result is obtained when the target ID numbers are matched and consistent, the second type of matching result is obtained when the target ID numbers are not matched but the target speed, distance and angle information are matched and consistent, and the third type of matching result is obtained when the target ID numbers, the target speed, the distance and the angle information are not matched; if the matching result is the first two types, the index position of the target on the radar distance-velocity diagram is reversely deduced according to the target track information obtained by tracking, and effective micro Doppler data are extracted according to the index position and the target track information obtained by current tracking is stored at the position of the current frame in the accumulation matrix; respectively marking the first class matching result and the second class matching result as a matching state 1 and a matching state 2; if the matching result is the first type, subsequent matching operation is not carried out, if the matching result is the second type, matching between target track information obtained by tracking the current target and next target track information in the accumulation matrix is carried out on the current matching result, if the matching result can be upgraded to the first type, the micro Doppler data and the target track information accumulated in the accumulation matrix are updated, and if the matching result cannot be upgraded to the first type, the original micro Doppler data and the target track information are maintained; if the matching result is of a third type, marking the target track as a matching state 3 and continuing to match the target track information obtained by tracking the current target with the next target track information in the accumulation matrix;

when the target track information obtained by tracking the current target is matched with all the target track information in the accumulation matrix, finishing the inner circulation, and turning to the outer circulation operation; in the outer loop operation, the micro-doppler data and the target track information are newly created for the target in the matching state 3, and the micro-doppler data and the target track information of the target are stored at a new position in the accumulation matrix for accumulation.

6. A pedestrian identification device based on tracking micro-doppler plots, comprising:

a memory storing a computer program;

a processor for executing the computer program, the computer program executing the steps of the method for identifying a pedestrian based on a tracking micro-doppler map according to any one of claims 1 to 5.

Technical Field

The invention relates to a target identification method, in particular to a pedestrian identification method based on a tracking micro Doppler image.

Background

With the development of economy and the progress of science and technology, people also gradually promote the safety demand, and the research of pedestrian recognition technology has important meaning to intelligent driving, traffic accident prevention, security protection field and the like. The pedestrian is identified through various sensors, so that the vehicle obtains signals to avoid the pedestrian, and the occurrence of traffic accidents is reduced; and aiming at the regional security protection requirement, the pedestrian invasion identification and early warning of the regional boundary can be carried out.

In the prior art, the algorithm for classifying targets based on the millimeter wave radar mainly comprises the following steps, some methods propose target feature extraction based on radar point cloud data, mainly comprise features such as broadening of targets in different directions, speed mean variance, mean variance of radar scattering cross sections and the like, and different support vector machines are adopted to complete target classification. However, such methods depend on the quality and quantity of point cloud data, have high requirements on hardware conditions of millimeter wave radars, and require radars to have high angular resolution. Still other methods use a single or combined multiple distance-doppler (RD) maps to extract features of target distance dimension broadening, velocity dimension broadening, etc. that can reflect the physical structure or micromotion of the target, and extract the differences of RD maps of previous and subsequent frames, and use a support vector machine to complete the classification of the target. This class of feature extraction methods is limited to RD graphs and is not well suited in multi-objective situations.

Disclosure of Invention

The invention aims to overcome the defects in the prior art, and provides a pedestrian identification method and device based on a tracking micro-Doppler image, which can eliminate environmental interference and improve identification accuracy, have the characteristics of high safety and multi-target applicability, can effectively detect the target position in real time, can also quickly identify pedestrians, and have higher applicability in engineering. In order to achieve the technical purpose, the embodiment of the invention adopts the technical scheme that:

in a first aspect, an embodiment of the present invention provides a pedestrian identification method based on a tracking micro-doppler map, including the following steps:

step S10, radar signal data is obtained;

step S20, processing the radar signal data to obtain radar point cloud data;

step S30, clustering and target tracking are carried out on the radar point cloud data to obtain target track information;

step S40, screening to obtain a stably existing target based on the target track information, then extracting micro Doppler data of the stably existing target, and accumulating multi-frame effective micro Doppler data to obtain a target micro Doppler image;

step S50, selecting target speed, distance, intensity and accumulated target micro Doppler image in the target track information as effective characteristics;

step S60, training a classifier according to effective characteristics obtained by a plurality of groups of radar signal data of different classes of targets;

in step S70, the features of the various types of objects are classified by the classifier, and the object having the pedestrian attribute is screened out.

Further, the radar signal data is radar intermediate frequency signal data obtained by processing radar echo signals after reflected radar detection signals are transmitted.

Further, the processing the radar signal data to obtain radar point cloud data specifically includes: and performing one-dimensional Fourier transform, two-dimensional Fourier transform, constant false alarm detection and digital beam forming angle measurement signal processing on the radar signal data to obtain radar point cloud data, wherein the radar point cloud data comprises target speed, distance, angle and intensity information.

Further, the screening to obtain the stably existing target based on the target track information specifically includes: and screening to obtain the stably existing target according to the existing length of the target track.

Further, the extracting of the micro doppler data of the stably existing target and accumulating the effective multi-frame micro doppler data to obtain the target micro doppler map specifically include:

step S401, initializing an accumulation matrix, wherein the accumulation matrix is used for storing accumulated micro Doppler data and target track information;

step S402, matching the tracked target track information with the target track information accumulated in the accumulation matrix, wherein the target track information specifically comprises inner circulation operation and outer circulation operation;

in the inner loop operation, matching the target track information obtained by tracking the ith target with the jth target track information in the accumulation matrix; the matching results are divided into three categories: the first type of matching result is obtained when the target ID numbers are matched and consistent, the second type of matching result is obtained when the target ID numbers are not matched but the target speed, distance and angle information are matched and consistent, and the third type of matching result is obtained when the target ID numbers, the target speed, the distance and the angle information are not matched; if the matching result is the first two types, the index position of the target on the radar distance-velocity diagram is reversely deduced according to the target track information obtained by tracking, and effective micro Doppler data are extracted according to the index position and the target track information obtained by current tracking is stored at the position of the current frame in the accumulation matrix; respectively marking the first class matching result and the second class matching result as a matching state 1 and a matching state 2; if the matching result is the first type, subsequent matching operation is not carried out, if the matching result is the second type, matching between target track information obtained by tracking the current target and next target track information in the accumulation matrix is carried out on the current matching result, if the matching result can be upgraded to the first type, the micro Doppler data and the target track information accumulated in the accumulation matrix are updated, and if the matching result cannot be upgraded to the first type, the original micro Doppler data and the target track information are maintained; if the matching result is of a third type, marking the target track as a matching state 3 and continuing to match the target track information obtained by tracking the current target with the next target track information in the accumulation matrix;

when the target track information obtained by tracking the current target is matched with all the target track information in the accumulation matrix, finishing the inner circulation, and turning to the outer circulation operation; in the outer loop operation, the micro-doppler data and the target track information are newly created for the target in the matching state 3, and the micro-doppler data and the target track information of the target are stored at a new position in the accumulation matrix for accumulation.

In a second aspect, an embodiment of the present invention provides a pedestrian identification apparatus based on a tracking micro-doppler map, including:

a memory storing a computer program;

a processor for executing the computer program, the computer program executing the steps of the method for identifying a pedestrian based on a tracking micro-doppler map as described above.

The technical scheme provided by the embodiment of the invention has the following beneficial effects:

1) only radar signal data are needed to be utilized, combination with other sensors is not needed, cost and calculation complexity are reduced, and usability is high.

2) The method can eliminate environmental interference and improve the identification accuracy.

3) Combining target feature extraction with a tracking algorithm, so that each target has a category attribute and can be matched with the tracked target track information; the method can realize the stable identification of multiple targets in the scene, and cannot influence the identification result due to the abnormity or loss of the single-frame radar signal.

Drawings

Fig. 1 is a schematic structural diagram of a millimeter wave radar in an embodiment of the present invention.

Fig. 2 is a flowchart of a pedestrian identification method in the embodiment of the invention.

Fig. 3 is a flowchart of an algorithm for obtaining a target micro-doppler map by accumulating micro-doppler data according to an embodiment of the present invention.

Detailed Description

In order to make the objects, technical solutions and advantages of the present invention more apparent, the present invention is described in further detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative of the invention and are not intended to limit the invention.

The embodiment of the invention takes a millimeter wave radar as an example, and introduces a pedestrian identification method and a device based on a tracking micro Doppler image; as shown in fig. 1, a millimeter wave radar includes a signal transmitting module, a signal receiving module, a frequency mixing module, and a processor; the signal transmitting module is used for transmitting radar detection signals to the vehicle, radar echo signals of the radar detection signals after being reflected by the target are received by the signal receiving module, and then radar intermediate frequency signal data are obtained through processing of the frequency mixing module; the processor acquires and processes the radar intermediate frequency signal data;

an embodiment of the present invention provides a pedestrian identification method based on a tracking micro-doppler map, as shown in fig. 2, including the following steps:

step S10, radar signal data is obtained; the radar signal data is radar intermediate frequency signal data obtained by processing radar echo signals after reflected radar detection signals are transmitted;

step S20, processing the radar signal data to obtain radar point cloud data;

step S30, clustering and target tracking are carried out on the radar point cloud data to obtain target track information;

step S40, screening to obtain a stably existing target based on the target track information, then extracting micro Doppler data of the stably existing target, and accumulating multi-frame effective micro Doppler data to obtain a target micro Doppler image;

step S50, selecting target speed, distance, intensity and accumulated target micro Doppler image in the target track information as effective characteristics;

step S60, training a classifier according to effective characteristics obtained by a plurality of groups of radar signal data of different classes of targets;

step S70, classifying the characteristics of various targets by using a classifier, and screening out the targets with pedestrian attributes;

in the embodiment of the invention, the target track information after clustering and tracking is more stable, some randomly occurring interference targets and noise signals can be filtered, and the position information of the current frame of the target and the position information of the next frame of the predicted target can be optimized; the classifier is trained by taking the target speed, distance and strength in the target track information and the target micro-Doppler image obtained by accumulation as effective features, so that the classifier can obtain a better classification effect, the pedestrian identification accuracy is improved, and different types of multi-target identification can be carried out;

specifically, in step S20, the processing the radar signal data to obtain radar point cloud data specifically includes: performing one-dimensional Fourier transform, two-dimensional Fourier transform, constant false alarm detection and digital beam forming (DBF for short) angle measurement signal processing on the radar signal data to obtain radar point cloud data, wherein the radar point cloud data comprises target speed, distance, angle and intensity information;

specifically, in step S40, the screening to obtain stably existing targets based on the target trajectory information specifically includes: screening according to the existence length of the target track to obtain a stably existing target; for example, whether the target track exists stably can be judged according to whether the existing length of the target track is greater than a set threshold;

as shown in fig. 3, in step S40, the extracting of micro doppler data from a stably existing target and accumulating effective micro doppler data of multiple frames to obtain a target micro doppler map specifically includes:

step S401, initializing an accumulation matrix, wherein the accumulation matrix is used for storing accumulated micro Doppler data and target track information;

step S402, matching the tracked target track information with the target track information accumulated in the accumulation matrix, wherein the target track information specifically comprises inner circulation operation and outer circulation operation;

in the inner loop operation, matching the target track information obtained by tracking the ith target with the jth target track information in the accumulation matrix; the matching results are divided into three categories: the first type of matching result is obtained when the target ID numbers are matched and consistent, the second type of matching result is obtained when the target ID numbers are not matched but the target speed, distance and angle information are matched and consistent, and the third type of matching result is obtained when the target ID numbers, the target speed, the distance and the angle information are not matched; if the matching result is the first two types, the index position of the target on the radar distance-velocity diagram is reversely deduced according to the target track information obtained by tracking, and effective micro Doppler data are extracted according to the index position and the target track information obtained by current tracking is stored at the position of the current frame in the accumulation matrix; respectively marking the first class matching result and the second class matching result as a matching state 1 and a matching state 2; if the matching result is the first type, subsequent matching operation is not carried out, if the matching result is the second type, matching between target track information obtained by tracking the current target and next target track information in the accumulation matrix is carried out on the current matching result, if the matching result can be upgraded to the first type, the micro Doppler data and the target track information accumulated in the accumulation matrix are updated, and if the matching result cannot be upgraded to the first type, the original micro Doppler data and the target track information are maintained; if the matching result is of a third type, marking the target track as a matching state 3 and continuing to match the target track information obtained by tracking the current target with the next target track information in the accumulation matrix;

when the target track information obtained by tracking the current target is matched with all the target track information in the accumulation matrix, finishing the inner circulation, and turning to the outer circulation operation; in the outer loop operation, the micro-doppler data and the target track information are newly created for the target in the matching state 3, and the micro-doppler data and the target track information of the target are stored at a new position in the accumulation matrix for accumulation.

In summary, the present application provides a pedestrian recognition method capable of eliminating environmental interference, improving recognition accuracy, having high security and being suitable for multiple targets, which not only can effectively detect the target position in real time, but also can quickly recognize pedestrians, and has better applicability in engineering.

Finally, it should be noted that the above embodiments are only for illustrating the technical solutions of the present invention and not for limiting, and although the present invention has been described in detail with reference to examples, it should be understood by those skilled in the art that modifications or equivalent substitutions may be made on the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention, which should be covered by the claims of the present invention.