CN113866742A - Method for point cloud processing and target classification of 4D millimeter wave radar - Google Patents
Method for point cloud processing and target classification of 4D millimeter wave radar Download PDFInfo
- Publication number
- CN113866742A CN113866742A CN202111466169.2A CN202111466169A CN113866742A CN 113866742 A CN113866742 A CN 113866742A CN 202111466169 A CN202111466169 A CN 202111466169A CN 113866742 A CN113866742 A CN 113866742A
- Authority
- CN
- China
- Prior art keywords
- point
- target
- trace
- track
- frame
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Granted
Links
Images
Classifications
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01S—RADIO DIRECTION-FINDING; RADIO NAVIGATION; DETERMINING DISTANCE OR VELOCITY BY USE OF RADIO WAVES; LOCATING OR PRESENCE-DETECTING BY USE OF THE REFLECTION OR RERADIATION OF RADIO WAVES; ANALOGOUS ARRANGEMENTS USING OTHER WAVES
- G01S7/00—Details of systems according to groups G01S13/00, G01S15/00, G01S17/00
- G01S7/02—Details of systems according to groups G01S13/00, G01S15/00, G01S17/00 of systems according to group G01S13/00
- G01S7/41—Details of systems according to groups G01S13/00, G01S15/00, G01S17/00 of systems according to group G01S13/00 using analysis of echo signal for target characterisation; Target signature; Target cross-section
- G01S7/415—Identification of targets based on measurements of movement associated with the target
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01S—RADIO DIRECTION-FINDING; RADIO NAVIGATION; DETERMINING DISTANCE OR VELOCITY BY USE OF RADIO WAVES; LOCATING OR PRESENCE-DETECTING BY USE OF THE REFLECTION OR RERADIATION OF RADIO WAVES; ANALOGOUS ARRANGEMENTS USING OTHER WAVES
- G01S7/00—Details of systems according to groups G01S13/00, G01S15/00, G01S17/00
- G01S7/02—Details of systems according to groups G01S13/00, G01S15/00, G01S17/00 of systems according to group G01S13/00
- G01S7/41—Details of systems according to groups G01S13/00, G01S15/00, G01S17/00 of systems according to group G01S13/00 using analysis of echo signal for target characterisation; Target signature; Target cross-section
- G01S7/418—Theoretical aspects
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01S—RADIO DIRECTION-FINDING; RADIO NAVIGATION; DETERMINING DISTANCE OR VELOCITY BY USE OF RADIO WAVES; LOCATING OR PRESENCE-DETECTING BY USE OF THE REFLECTION OR RERADIATION OF RADIO WAVES; ANALOGOUS ARRANGEMENTS USING OTHER WAVES
- G01S13/00—Systems using the reflection or reradiation of radio waves, e.g. radar systems; Analogous systems using reflection or reradiation of waves whose nature or wavelength is irrelevant or unspecified
- G01S13/02—Systems using reflection of radio waves, e.g. primary radar systems; Analogous systems
- G01S13/06—Systems determining position data of a target
- G01S13/42—Simultaneous measurement of distance and other co-ordinates
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01S—RADIO DIRECTION-FINDING; RADIO NAVIGATION; DETERMINING DISTANCE OR VELOCITY BY USE OF RADIO WAVES; LOCATING OR PRESENCE-DETECTING BY USE OF THE REFLECTION OR RERADIATION OF RADIO WAVES; ANALOGOUS ARRANGEMENTS USING OTHER WAVES
- G01S13/00—Systems using the reflection or reradiation of radio waves, e.g. radar systems; Analogous systems using reflection or reradiation of waves whose nature or wavelength is irrelevant or unspecified
- G01S13/02—Systems using reflection of radio waves, e.g. primary radar systems; Analogous systems
- G01S13/50—Systems of measurement based on relative movement of target
- G01S13/58—Velocity or trajectory determination systems; Sense-of-movement determination systems
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01S—RADIO DIRECTION-FINDING; RADIO NAVIGATION; DETERMINING DISTANCE OR VELOCITY BY USE OF RADIO WAVES; LOCATING OR PRESENCE-DETECTING BY USE OF THE REFLECTION OR RERADIATION OF RADIO WAVES; ANALOGOUS ARRANGEMENTS USING OTHER WAVES
- G01S13/00—Systems using the reflection or reradiation of radio waves, e.g. radar systems; Analogous systems using reflection or reradiation of waves whose nature or wavelength is irrelevant or unspecified
- G01S13/66—Radar-tracking systems; Analogous systems
- G01S13/72—Radar-tracking systems; Analogous systems for two-dimensional tracking, e.g. combination of angle and range tracking, track-while-scan radar
- G01S13/723—Radar-tracking systems; Analogous systems for two-dimensional tracking, e.g. combination of angle and range tracking, track-while-scan radar by using numerical data
- G01S13/726—Multiple target tracking
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01S—RADIO DIRECTION-FINDING; RADIO NAVIGATION; DETERMINING DISTANCE OR VELOCITY BY USE OF RADIO WAVES; LOCATING OR PRESENCE-DETECTING BY USE OF THE REFLECTION OR RERADIATION OF RADIO WAVES; ANALOGOUS ARRANGEMENTS USING OTHER WAVES
- G01S13/00—Systems using the reflection or reradiation of radio waves, e.g. radar systems; Analogous systems using reflection or reradiation of waves whose nature or wavelength is irrelevant or unspecified
- G01S13/88—Radar or analogous systems specially adapted for specific applications
- G01S13/93—Radar or analogous systems specially adapted for specific applications for anti-collision purposes
- G01S13/931—Radar or analogous systems specially adapted for specific applications for anti-collision purposes of land vehicles
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01S—RADIO DIRECTION-FINDING; RADIO NAVIGATION; DETERMINING DISTANCE OR VELOCITY BY USE OF RADIO WAVES; LOCATING OR PRESENCE-DETECTING BY USE OF THE REFLECTION OR RERADIATION OF RADIO WAVES; ANALOGOUS ARRANGEMENTS USING OTHER WAVES
- G01S7/00—Details of systems according to groups G01S13/00, G01S15/00, G01S17/00
- G01S7/02—Details of systems according to groups G01S13/00, G01S15/00, G01S17/00 of systems according to group G01S13/00
- G01S7/41—Details of systems according to groups G01S13/00, G01S15/00, G01S17/00 of systems according to group G01S13/00 using analysis of echo signal for target characterisation; Target signature; Target cross-section
- G01S7/414—Discriminating targets with respect to background clutter
Landscapes
- Engineering & Computer Science (AREA)
- Radar, Positioning & Navigation (AREA)
- Remote Sensing (AREA)
- Physics & Mathematics (AREA)
- Computer Networks & Wireless Communication (AREA)
- General Physics & Mathematics (AREA)
- Electromagnetism (AREA)
- Radar Systems Or Details Thereof (AREA)
Abstract
The invention discloses a point cloud processing and target classification method for a 4D millimeter wave radar. The method comprises the steps of trace point input, trace point preprocessing, Kalman filtering prediction, association of trace points and tracks, clustering of trace points, starting of tracks, updating of tracks and track management. The invention realizes the transformation from a two-dimensional plane to a three-dimensional plane, and the trace point characteristic of the target is more obvious. The method comprises the steps of constructing a target measurement virtual point trace in a signal-to-noise ratio weighting mode, projecting an associated point trace on an xoy plane, rotating a course angle of the xoy plane clockwise by taking the virtual point trace as an original point, calculating size information of a target by adopting a multi-frame sliding window mode and a virtual point position relation displacement among multiple frames, improving the problem that the size of the target is not obvious due to the fact that millimeter wave point cloud is sparse, obtaining the classification probability of the target of a frame according to the class probability given to the target of the single frame according to the characteristics of the target, combining a historical probability and a single frame probability weighting mode, and adopting the class with the largest frame probability as the final classification result of the frame.
Description
Technical Field
The invention relates to the technical field of 4D millimeter wave radar point cloud processing and target classification methods, in particular to a 4D millimeter wave radar point cloud processing and target classification method.
Background
At present, point clouds of millimeter wave radars are sparse, the contained target features are few, and the influence of different directions of targets on target classification in the driving process is not fully considered in a classification method using point trace features, so that the target classification method based on the millimeter wave radars is low in accuracy and poor in practicability, great challenges are brought to the development of the millimeter wave radars in practical application, the existing target classification based on the millimeter wave radars is mainly applied to distinguishing pedestrians and vehicles, and the market demands for target classification are not only the same. The 4D millimeter wave radar increases the pitch angle information on the original distance, horizontal angle and speed, and expands a target from a two-dimensional plane to a three-dimensional space in space, so that the shape characteristic of the target is more obvious. Under the conditions that point cloud of a traditional millimeter wave radar is sparse and various targets influence target classification judgment in reality at different azimuth angles, the 4D radar further improves the number and quality of traces, meanwhile, point cloud processing is carried out on a three-dimensional space to extract course information of the targets, the trace rotation is carried out according to the course angle information by associating the trace information with the targets through a multi-frame sliding window, the length, width and height shape characteristics of the targets are calculated through trace displacement based on the position relation of multi-frame virtual traces, the unobvious characteristics of the sizes of the targets are improved, and the applicability of the classification of the targets in different azimuth driving is improved. Meanwhile, the length, width, height, RCS, volume and the like of the target are used as the characteristics of the target, the probability of each class of the single-frame target is given, the classification probability of the target of the frame is obtained by combining the historical probability and the single-frame probability weighting mode, the class with the maximum probability of the frame is used as the final classification result of the frame, pedestrians, two-wheel vehicles, trolleys and commercial vehicles can be distinguished in real time, and the accuracy and universality of target classification are further improved.
Disclosure of Invention
The invention aims to provide a method for processing point cloud and classifying targets of a 4D millimeter wave radar aiming at the defects in the prior art.
In order to achieve the above object, the present invention provides a method for point cloud processing and target classification of a 4D millimeter wave radar, comprising:
inputting a point trace of a target acquired by a 4D millimeter wave radar, and preprocessing the point trace, wherein the measurement of the point trace comprises the distance of the targetHorizontal anglePitch angleAnd radial velocity;
Performing Kalman filtering prediction on all existing tracks, and then predicting the loss times of all tracks after predictionAnd life cycle of all tracksRespectively adding 1 to the extrapolated time of the flight path and the radar period T;
associating the preprocessed trace points with the existing flight path before the frame;
clustering the preprocessed point tracks which are not related to the upper track by adopting a density clustering mode;
carrying out initial track on the clustering result;
and updating the flight path, wherein the flight path updating comprises Kalman filtering updating, and the Kalman filtering updating mode is as follows: traversing all the effective tracks, judging whether each track has an associated point track in the frame, if so, initializing a Kalman filter for the track, and if the track is more than two frames, updating Kalman filtering; then, the following operations are carried out on all tracks with associated point tracks in the frame: updating the radar scattering cross section of the flight path to be the maximum value of the radar scattering cross section in the associated point path of the frame, resetting the number of the associated point paths of the flight path to be 0, and losing the flight pathSubtracting 1, and resetting the extrapolation time of the flight path to 0;
and deleting the flight path of the unreal target or the flight path which cannot be stably tracked when the target is not in the radar detection range.
Further, the pretreatment comprises angle correction, and the horizontal angle after the angle correctionPitch angleRespectively as follows:
wherein the content of the first and second substances,respectively the level angle and the pitch installation angle of the calibrated radar.
Further, the pretreatment also comprises dynamic and static separation, and the dynamic and static separation mode is as follows:
radial velocity of the spotDecompose intoPlane velocity of the objectDecomposing the speed of the vehicle where the radar is located into a point track and projecting the point track in the radial directionPlane velocity of the objectRadial velocity of the spotResolving to Z-axis direction to obtain velocityIf velocityAnd speedSum less than thresholdAnd speed ofLess than thresholdIf not, the trace point is judged as a moving target.
Further, associating the preprocessed trace point with the existing track before the current frame specifically includes: when the preprocessed point track passes through a distance wave gate, a horizontal angle wave gate, a pitching angle wave gate and a radial speed wave gate set by a certain track, the distance between the point track and the track is recorded, and finally, the nearest track is associated with the predicted point track by adopting a nearest neighbor association mode.
Further, the step of starting the track of the clustering result specifically includes:
traversing the storage of the clustering result and the flight path, and if a certain position in the array for storing the flight path is empty, storing a new flight path at the position;
and (3) performing dynamic and static inspection on the traces in each class, specifically as follows: if the number of the motion points exceeds the total associated point trace ratio threshold valueMarking the track as non-stationary, otherwise, marking the track as stationary;
constructing a virtual measuring point track of the initial flight path, which comprises the following specific steps:
wherein the content of the first and second substances,the number of all traces in a class,to construct the distance of the virtual metrology trace,the distance of the ith trace point within the class,to construct the horizontal angle of the virtual measurement trace,is the horizontal angle of the ith point trace in the class,To construct the pitch angle of the virtual metrology trace,the pitch angle of the ith point trace in the class,to construct the radial velocity of the virtual metrology trace,the radial velocity of the ith point trace in the class,the sum weight of the signal-to-noise ratios of all points in the signal-to-noise ratio occupation class of each trace is taken as the weight of the sum of the signal-to-noise ratios of all points in the signal-to-noise ratio occupation class of each trace;
of the initial flight pathGenerating a periodInitialized to 1, number of lossesThe initialization is 0, the filtering state is set to be uninitialized, the track associated point track number is initialized to be 0, and the distance of the initial track isHorizontal angle ofA pitch angle ofAnd a radial velocity of。
Further, the track update further includes a target classification, and the target classification is performed in the following manner:
the sliding window stores the track information of a plurality of frames of targets, and the track information is included inCalculating the maximum values of projection points of all associated point tracks of each frame of flight path on the xoy surface after clockwise rotating according to the course angle alpha and then respectively rotating on the X axis and the Y axis、And the minimum value of projection points of all associated point tracks of each frame track on the xoy surface on the X axis and the Y axis after clockwise rotation according to the heading angle alpha、Calculating the maximum and minimum values of all associated point tracks of each frame track on the Z axis、;
Respectively calculating multi-frame track associated point tracks、、Andrespectively carrying out displacement according to the coincidence of the virtual measuring point trace coordinate of each frame and the measuring point trace coordinate of the first frame, and then respectively obtaining the maximum values on the X axis and the Y axis after the displacementAndand minimum values in X-axis and Y-axis after translation, respectivelyAnd(ii) a Of multiple frames、Respectively carrying out displacement according to the coincidence of the Z-axis coordinate of the virtual measuring point trace of each frame and the Z-axis coordinate of the measuring point trace of the first frame, and taking the maximum value after displacementAnd minimum valueRespectively calculating:
Giving the length of each class, the radar scattering cross section and the corresponding threshold value of the volume according to the distance and the horizontal angle of the target, and determining the length of the targetSpeed, velocityCalculating the class probability of a single-frame target based on the characteristics of the target, calculating the class probability of the single-frame target in a weighted form by combining the historical probability and the class probability of the single-frame target, and taking the class with the maximum class probability of the frame target as the final classification result of the frame, wherein,
、the velocities on the x-axis and the y-axis after the k-th kalman filter update of the target, respectively, k being an integer greater than zero.
Has the advantages that: the target tracking method based on the 4D millimeter wave point cloud data realizes the conversion from a two-dimensional plane to a three-dimensional plane, and the point trace characteristics of the target are more obvious. On the basis of target tracking, a course angle alpha on an xoy plane is calculated through a filtering speed of a moving track, a target measurement virtual point track is constructed in a signal-to-noise ratio weighting mode, an associated point track is projected onto the xoy plane, the virtual point track serves as an original point and rotates alpha clockwise, meanwhile, the size information of a target is calculated in a multi-frame sliding window mode through the displacement of the virtual point position relation among multiple frames, the problem that the size of the target is not obvious due to the fact that millimeter wave point cloud is sparse is solved, meanwhile, the length, RCS, the volume and the like of the target serve as the characteristics of the target, the class probability of a single-frame target is given, the classification probability of the frame target is obtained through combination of a historical probability and a single-frame probability weighting mode, the class with the largest probability of the frame serves as the final classification result of the frame, and classification of people, two-wheeled vehicles, vehicles and commercial vehicles is achieved. The invention meets the future intelligent requirement of the development of the vehicle-mounted millimeter wave radar and has important practical significance on the research of the automatic driving perception capability.
Drawings
FIG. 1 is a schematic diagram of a coordinate system employed for trace point processing;
FIG. 2 is a flow chart of a method of 4D millimeter wave radar point cloud processing and target classification in an embodiment of the invention;
FIG. 3 is a flow chart of trace point clustering according to an embodiment of the invention;
FIG. 4 is a flow diagram illustrating object classification according to an embodiment of the present invention;
FIG. 5 is a schematic diagram of a sliding window storing multiple frames of target information;
FIG. 6 is a schematic diagram of calculating the length, width and height of a target;
FIG. 7 is a flow diagram of classification based on calculating class probabilities.
Detailed Description
The present invention will be further illustrated with reference to the accompanying drawings and specific examples, which are carried out on the premise of the technical solution of the present invention, and it should be understood that these examples are only for illustrating the present invention and are not intended to limit the scope of the present invention.
As shown in fig. 1 to 7, an embodiment of the present invention provides a method for point cloud processing and target classification of a 4D millimeter wave radar, which performs point trace processing based on the coordinate system shown in fig. 1, where a radar is installed at a coordinate dot in fig. 1, and fig. 1 is a diagram where the radar is installed correctly, that is, a normal line coincides with an X axis. The method specifically comprises the following steps:
inputting a point trace of a target acquired by a 4D millimeter wave radar, preprocessing the point trace, and measuring the point trace to obtain the distance of the targetHorizontal anglePitch angleAnd radial velocity. The preprocessing of the trace comprises angle correction and horizontal angle after the angle correctionPitch angleRespectively as follows:
wherein the content of the first and second substances,respectively the level angle and the pitch installation angle of the calibrated radar.
The pretreatment of the trace also preferably comprises dynamic and static separation, and the dynamic and static separation mode is as follows:
radial velocity of the spotDecompose intoPlane velocity of the objectDecomposing the speed of the vehicle where the radar is located into a point track and projecting the point track in the radial directionPlane velocity of the objectRadial velocity of the spotResolving to Z-axis direction to obtain velocityIf velocityAnd speedSum less than thresholdAnd speed ofLess than thresholdIf not, the trace point is judged as a moving target.
Performing Kalman filtering prediction on all existing tracks, calculating prediction covariance, and the distance, horizontal angle, pitch angle and radial speed of the predicted target, and predicting the loss times of all tracksAnd life cycle of all tracks1 is added to each, and the extrapolated time of the flight path is added to the radar period T. Specifically, the Kalman filtering prediction is carried out by utilizing a uniform linear motion model, and an equation of the uniform linear motion model is as follows:
wherein the content of the first and second substances,to predict the state vector of the acquired target measurement point trace for the (k + 1) th measurement point,in order to be a noise of the process,to predict the state vector of the target kth measurement point trace obtained,can be expressed as:
wherein the content of the first and second substances,、、、、andrespectively sequentially updating the position on an x axis, the position on a y axis, the position on a z axis, the speed on the x axis, the speed on the y axis and the speed on the z axis of the target after the k-th filtering, wherein k is an integer greater than 0;
wherein T is radar single frame time.
、、andrespectively the distance of the k-th measuring point trace of the targetHorizontal angle, pitch angle and radial velocity, then its observation equation is:
wherein the content of the first and second substances,in order to measure the noise, the noise is measured,for the observation function:
And associating the preprocessed point track with the existing track before the current frame. The method comprises the following specific steps: when the preprocessed point track passes through a distance wave gate, a horizontal angle wave gate, a pitching angle wave gate and a radial speed wave gate set by a certain track, the distance between the point track and the track is recorded, and finally, the nearest track is associated with the predicted point track by adopting a nearest neighbor association mode. The principle of associating the trace points with the tracks is that a single trace point can only be associated with a single track, but a single track can be associated with a plurality of traces.
Referring to fig. 3, the embodiment of the present invention preferably uses Density-Based Spatial Clustering of Applications with Noise (DBSCAN) to cluster predicted traces that are not associated with tracks. Specifically, set variables、If the Euclidean distance between trace A and trace B is smaller than the Euclidean distanceIf the number of trace points in the neighborhood of trace point A is less than the number of trace points in the neighborhood of trace point A, then trace point B is called the trace point in the neighborhood of trace point AThen, point trace A is called as a core object, and the points in the neighborhood of A are all reachable by the direct density of point A.
And carrying out initial track on the clustering result. Specifically, the clustering result and the flight path are stored, and if a certain position in the array for storing the flight path is empty, a new flight path is stored at the position.
And (3) performing dynamic and static inspection on the traces in each class, specifically as follows: if the number of the motion points exceeds the total associated point trace ratio threshold valueThe track is marked as non-stationary, otherwise, the track is marked as stationary.
Constructing a measuring point trace of the initial flight trace, wherein the signal-to-noise ratio of the point trace is higher than the intensity of the reflective signal, and the accuracy of the point trace in all aspects is generally considered to be higher if the signal-to-noise ratio is higher, so that a virtual measuring point trace is preferably constructed in a signal-to-noise ratio weighting manner, which is specifically as follows:
wherein the content of the first and second substances,the number of all traces in a class,to construct the distance of the virtual metrology trace,the distance of the ith trace point within the class,to construct the horizontal angle of the virtual measurement trace,is the horizontal angle of the ith point trace in the class,To construct the pitch angle of the virtual metrology trace,the pitch angle of the ith point trace in the class,to construct the radial velocity of the virtual metrology trace,the radial velocity of the ith point trace in the class,the sum weight of the signal-to-noise ratios of all points in the signal-to-noise ratio occupation class of each trace is
Of the initial flight pathGenerating cycleNumber of times of lossThe filtering state is set to be not initialized, the track number of the track associated points is initialized to be 0, and the distance of the initial track isHorizontal angle ofA pitch angle ofAnd a radial velocity of。
And updating the flight path, wherein the updating of the flight path comprises Kalman filtering updating, specifically, traversing all effective flight paths, judging whether each flight path has an associated point path in the frame, if so, initializing a Kalman filter for the flight path, and if the flight path is more than two frames, updating the Kalman filtering. Thereby updating the track filter state and the filter covariance. Then, the following operations are carried out on all tracks with associated point tracks in the frame: updating the radar scattering cross section (RCS) of the flight path to the maximum value of the radar scattering cross section in the associated point path of the frame, resetting the number of the associated point paths of the flight path to 0 and the loss times of the flight pathSubtract 1 and reset the extrapolated time of the flight path to 0.
And carrying out track management to delete the track of the unreal target or the track which cannot be stably tracked when the target is not in the radar detection range. Specifically, all tracks are traversed, loss times and extrapolation time detection are carried out on effective tracks, and if the tracks are around L frames before the tracks in the track initialization stage and the track loss times are larger than or equal to m, the tracks are deleted. And if the extrapolation time of the flight path is greater than the radar cycle time of the s frames, deleting the flight path. In general, L can be from 4 to 8, m can be from 2 to 4, and s can be from 5 to 8.
The track update of the embodiment of the present invention further includes a target classification, see fig. 4, the target classification mode is as follows:
the sliding window stores the track information of a plurality of frames of targets, and the track information is included inCalculating the maximum values of projection points of all associated point tracks of each frame of flight path on the xoy surface after clockwise rotating according to the course angle alpha and then respectively rotating on the X axis and the Y axis、And the minimum value of projection points of all associated point tracks of each frame track on the xoy surface on the X axis and the Y axis after clockwise rotation according to the heading angle alpha、. Specifically, referring to fig. 5, the black dots are projected to the associated trace of the flight pathProjecting virtual measuring point trace with plane point and small grey dots as flight trace toPoints of the plane, alpha being according to the flight pathCalculating course angle by using speed on the plane, rotating all black small dots clockwise by alpha by using gray small dots as circle centers, and calculating the positions of the dots after the rotation is finishedX、YMaximum and minimum on axis、、And. Calculating the maximum and minimum values of all associated point tracks of each frame of flight path on the Z axis、,Anddirectly according to the maximum and minimum height of the associated trace point.
Referring to FIG. 6, the black origin in the graph is the projection of the first frame measurement trace ontoPoints of a plane, white origin being projected onto a measurement trace of some non-first frameThe virtual measuring point trace constructed by each frame is projected to the plane point and the gray round pointA point of the plane. Respectively calculating multi-frame track associated point tracks、、Andrespectively carrying out displacement according to the coincidence of the virtual measuring point trace coordinate of each frame and the measuring point trace coordinate of the first frame, and then respectively obtaining the maximum values on the X axis and the Y axis after the displacementAndand minimum values in X-axis and Y-axis after translation, respectivelyAnd(ii) a Of multiple frames、Respectively carrying out displacement according to the coincidence of the Z-axis coordinate of the virtual measuring point trace of each frame and the Z-axis coordinate of the measuring point trace of the first frame, and taking the maximum value after displacementAnd minimum valueRespectively calculating:
Referring to fig. 7, the length of the target is calculated according to the empirical values of each class obtained by the offline analysis, and the length, the radar cross-sectional area, and the threshold corresponding to the volume are given to each class according to the distance and the horizontal angle of the targetSpeed, velocityCalculating the class probability of a single-frame target based on the characteristics of the target, calculating the classification probability of the target in the frame in a weighted form by combining the historical probability and the class probability of the single-frame target, and taking the class with the maximum classification probability of the target in the frame as the final classification result of the frame, wherein,
、respectively the k-th Karman of the targetThe velocity on the x-axis and on the y-axis after the filter update, k being an integer greater than zero. The above categories include pedestrians, motorcycles, cars, and commercial vehicles (trucks, buses), etc.
The foregoing is only a preferred embodiment of the present invention, and it should be noted that other parts not specifically described are within the prior art or common general knowledge to those of ordinary skill in the art. Without departing from the principle of the invention, several improvements and modifications can be made, and these improvements and modifications should also be construed as the scope of the invention.
Claims (6)
1. A method for point cloud processing and target classification of a 4D millimeter wave radar is characterized by comprising the following steps:
inputting a point trace of a target acquired by a 4D millimeter wave radar, and preprocessing the point trace, wherein the measurement of the point trace comprises the distance of the targetHorizontal anglePitch angleAnd radial velocity;
Performing Kalman filtering prediction on all existing tracks, and then predicting the loss times of all tracks after predictionAnd life cycle of all tracksRespectively adding 1 to the extrapolated time of the flight path and the radar period T;
associating the preprocessed trace points with the existing flight tracks;
clustering the preprocessed point tracks which are not related to the upper track by adopting a density clustering mode;
carrying out initial track on the clustering result;
and updating the flight path, wherein the flight path updating comprises Kalman filtering updating, and the Kalman filtering updating mode is as follows: traversing all the effective tracks, judging whether each track has an associated point track in the frame, if so, initializing a Kalman filter for the track, and if the track is more than two frames, updating Kalman filtering; then, the following operations are carried out on all tracks with associated point tracks in the frame: updating the radar scattering cross section of the flight path to be the maximum value of the radar scattering cross section in the associated point path of the frame, resetting the number of the associated point paths of the flight path to be 0, and losing the flight pathSubtracting 1, and resetting the extrapolation time of the flight path to 0;
and deleting the flight path of the unreal target or the flight path which cannot be stably tracked when the target is not in the radar detection range.
2. The method for 4D millimeter wave radar point cloud processing and target classification as claimed in claim 1, wherein the pre-processing comprises angle rectification, horizontal angle after angle rectificationPitch angleRespectively as follows:
3. The method for point cloud processing and target classification of a 4D millimeter wave radar according to claim 1, wherein the pre-processing further comprises dynamic and static separation in the following manner:
radial velocity of the spotDecompose intoPlane velocity of the objectDecomposing the speed of the vehicle where the radar is located into a point track and projecting the point track in the radial directionPlane velocity of the objectRadial velocity of the spotResolving to Z-axis direction to obtain velocityIf velocityAnd speedSum less than thresholdAnd speed ofLess than thresholdIf not, the trace point is judged as a moving target.
4. The method for point cloud processing and target classification of a 4D millimeter wave radar according to claim 1, wherein associating the preprocessed point trace with an existing track of the current frame specifically comprises: when the preprocessed point track passes through a distance wave gate, a horizontal angle wave gate, a pitching angle wave gate and a radial speed wave gate set by a certain track, the distance between the point track and the track is recorded, and finally, the nearest track is associated with the predicted point track by adopting a nearest neighbor association mode.
5. The method for point cloud processing and target classification for 4D millimeter wave radar according to claim 1, wherein the step of starting the track for the result of clustering specifically comprises:
traversing the storage of the clustering result and the flight path, and if a certain position in the array for storing the flight path is empty, storing a new flight path at the position;
and (3) performing dynamic and static inspection on the traces in each class, specifically as follows: if the number of the motion points exceeds the total associated point trace ratio threshold valueMarking the track as non-stationary, otherwise, marking the track as stationary;
constructing a virtual measuring point track of the initial flight path, which comprises the following specific steps:
wherein the content of the first and second substances,the number of all traces in a class,to construct the distance of the virtual metrology trace,the distance of the ith trace point within the class,to construct the horizontal angle of the virtual measurement trace,is the horizontal angle of the ith point trace in the class,To construct the pitch angle of the virtual metrology trace,the pitch angle of the ith point trace in the class,to construct the radial velocity of the virtual metrology trace,the radial velocity of the ith point trace in the class,the sum weight of the signal-to-noise ratios of all points in the signal-to-noise ratio occupation class of each trace is taken as the weight of the sum of the signal-to-noise ratios of all points in the signal-to-noise ratio occupation class of each trace;
of the initial flight pathGenerating a periodInitialized to 1, number of lossesThe initialization is 0, the filtering state is set to be uninitialized, the track associated point track number is initialized to be 0, and the distance of the initial track isHorizontal angle ofA pitch angle ofAnd a radial velocity of。
6. The method of 4D millimeter wave radar point cloud processing and target classification of claim 1, wherein the track update further comprises a target classification by:
the sliding window stores the track information of a plurality of frames of targets, and the track information is included inCalculating the maximum values of projection points of all associated point tracks of each frame of flight path on the xoy surface after clockwise rotating according to the course angle alpha and then respectively rotating on the X axis and the Y axis、And the minimum value of projection points of all associated point tracks of each frame track on the xoy surface on the X axis and the Y axis after clockwise rotation according to the heading angle alpha、Calculating the maximum and minimum values of all associated point tracks of each frame track on the Z axis、;
Respectively calculating multi-frame track associated point tracks、、Andrespectively carrying out displacement according to the coincidence of the virtual measuring point trace coordinate of each frame and the measuring point trace coordinate of the first frame, and then respectively obtaining the maximum values on the X axis and the Y axis after the displacementAndand minimum values in X-axis and Y-axis after translation, respectivelyAnd(ii) a Of multiple frames、Respectively carrying out displacement according to the coincidence of the Z-axis coordinate of the virtual measuring point trace of each frame and the Z-axis coordinate of the measuring point trace of the first frame, and taking the maximum value after displacementAnd minimum valueRespectively calculating:
Giving the length of each class, the radar scattering cross section and the corresponding threshold value of the volume according to the distance and the horizontal angle of the target, and determining the length of the targetSpeed, velocityCalculating the class probability of a single-frame target based on the characteristics of the target, calculating the class probability of the single-frame target in a weighted form by combining the historical probability and the class probability of the single-frame target, and taking the class with the maximum class probability of the frame target as the final classification result of the frame, wherein,
Priority Applications (3)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202111466169.2A CN113866742B (en) | 2021-12-03 | 2021-12-03 | Method for point cloud processing and target classification of 4D millimeter wave radar |
PCT/CN2022/092002 WO2023097971A1 (en) | 2021-12-03 | 2022-05-10 | 4d millimeter wave radar data processing method |
DE112022000017.1T DE112022000017T5 (en) | 2021-12-03 | 2022-05-10 | DATA PROCESSING METHODS FOR 4D MILLIMETER WAVE RADAR |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202111466169.2A CN113866742B (en) | 2021-12-03 | 2021-12-03 | Method for point cloud processing and target classification of 4D millimeter wave radar |
Publications (2)
Publication Number | Publication Date |
---|---|
CN113866742A true CN113866742A (en) | 2021-12-31 |
CN113866742B CN113866742B (en) | 2022-02-22 |
Family
ID=78985803
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN202111466169.2A Active CN113866742B (en) | 2021-12-03 | 2021-12-03 | Method for point cloud processing and target classification of 4D millimeter wave radar |
Country Status (3)
Country | Link |
---|---|
CN (1) | CN113866742B (en) |
DE (1) | DE112022000017T5 (en) |
WO (1) | WO2023097971A1 (en) |
Cited By (10)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN115236674A (en) * | 2022-06-15 | 2022-10-25 | 北京踏歌智行科技有限公司 | Mining area environment sensing method based on 4D millimeter wave radar |
CN115656962A (en) * | 2022-12-26 | 2023-01-31 | 南京楚航科技有限公司 | Method for identifying height-limited object based on millimeter wave radar |
CN115825912A (en) * | 2023-01-09 | 2023-03-21 | 南京隼眼电子科技有限公司 | Radar signal processing method and device and storage medium |
CN115840221A (en) * | 2023-02-20 | 2023-03-24 | 上海几何伙伴智能驾驶有限公司 | Method for realizing target feature extraction and multi-target tracking based on 4D millimeter wave radar |
WO2023097971A1 (en) * | 2021-12-03 | 2023-06-08 | 南京楚航科技有限公司 | 4d millimeter wave radar data processing method |
CN116593973A (en) * | 2023-03-29 | 2023-08-15 | 深圳承泰科技有限公司 | Method and system for automatically calibrating installation angle of vehicle-mounted millimeter wave radar |
CN116990773A (en) * | 2023-09-27 | 2023-11-03 | 广州辰创科技发展有限公司 | Low-speed small target detection method and device based on self-adaptive threshold and storage medium |
CN117250595A (en) * | 2023-11-20 | 2023-12-19 | 长沙莫之比智能科技有限公司 | False alarm suppression method for vehicle-mounted millimeter wave radar metal well lid target |
CN117491965A (en) * | 2024-01-02 | 2024-02-02 | 上海几何伙伴智能驾驶有限公司 | Target track starting method based on 4D millimeter wave radar |
CN117647807A (en) * | 2024-01-30 | 2024-03-05 | 安徽隼波科技有限公司 | Motor vehicle size estimation method based on millimeter wave radar |
Families Citing this family (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN116881385B (en) * | 2023-09-08 | 2023-12-01 | 中国铁塔股份有限公司 | Track smoothing method, track smoothing device, electronic equipment and readable storage medium |
CN117647806B (en) * | 2024-01-30 | 2024-04-12 | 安徽隼波科技有限公司 | Point trace condensation and target tracking method based on millimeter wave radar |
Citations (7)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN110031834A (en) * | 2018-01-12 | 2019-07-19 | 西安艾索信息技术有限公司 | A kind of improved multiple target radar track processing method |
US20190302252A1 (en) * | 2018-03-27 | 2019-10-03 | Infineon Technologies Ag | System and method of monitoring an air flow using a millimeter-wave radar sensor |
CN111428573A (en) * | 2020-03-02 | 2020-07-17 | 南京莱斯电子设备有限公司 | Infrared weak and small target detection false alarm suppression method under complex background |
CN111929655A (en) * | 2020-09-08 | 2020-11-13 | 中国电子科技集团公司第三十八研究所 | Automobile millimeter wave radar road target tracking method and system |
CN112166336A (en) * | 2019-09-27 | 2021-01-01 | 深圳市大疆创新科技有限公司 | Method and device for calibrating pitching installation angle of millimeter wave radar, vehicle control system and vehicle |
CN113671481A (en) * | 2021-07-21 | 2021-11-19 | 西安电子科技大学 | 3D multi-target tracking processing method based on millimeter wave radar |
CN113721234A (en) * | 2021-08-30 | 2021-11-30 | 南京慧尔视智能科技有限公司 | Vehicle-mounted millimeter wave radar point cloud data dynamic and static separation filtering method and device |
Family Cites Families (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
WO2019234795A1 (en) * | 2018-06-04 | 2019-12-12 | 三菱電機株式会社 | Light impingement device |
CN113866742B (en) * | 2021-12-03 | 2022-02-22 | 南京楚航科技有限公司 | Method for point cloud processing and target classification of 4D millimeter wave radar |
-
2021
- 2021-12-03 CN CN202111466169.2A patent/CN113866742B/en active Active
-
2022
- 2022-05-10 WO PCT/CN2022/092002 patent/WO2023097971A1/en unknown
- 2022-05-10 DE DE112022000017.1T patent/DE112022000017T5/en active Pending
Patent Citations (7)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN110031834A (en) * | 2018-01-12 | 2019-07-19 | 西安艾索信息技术有限公司 | A kind of improved multiple target radar track processing method |
US20190302252A1 (en) * | 2018-03-27 | 2019-10-03 | Infineon Technologies Ag | System and method of monitoring an air flow using a millimeter-wave radar sensor |
CN112166336A (en) * | 2019-09-27 | 2021-01-01 | 深圳市大疆创新科技有限公司 | Method and device for calibrating pitching installation angle of millimeter wave radar, vehicle control system and vehicle |
CN111428573A (en) * | 2020-03-02 | 2020-07-17 | 南京莱斯电子设备有限公司 | Infrared weak and small target detection false alarm suppression method under complex background |
CN111929655A (en) * | 2020-09-08 | 2020-11-13 | 中国电子科技集团公司第三十八研究所 | Automobile millimeter wave radar road target tracking method and system |
CN113671481A (en) * | 2021-07-21 | 2021-11-19 | 西安电子科技大学 | 3D multi-target tracking processing method based on millimeter wave radar |
CN113721234A (en) * | 2021-08-30 | 2021-11-30 | 南京慧尔视智能科技有限公司 | Vehicle-mounted millimeter wave radar point cloud data dynamic and static separation filtering method and device |
Cited By (15)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
WO2023097971A1 (en) * | 2021-12-03 | 2023-06-08 | 南京楚航科技有限公司 | 4d millimeter wave radar data processing method |
CN115236674B (en) * | 2022-06-15 | 2024-06-04 | 北京踏歌智行科技有限公司 | Mining area environment sensing method based on 4D millimeter wave radar |
CN115236674A (en) * | 2022-06-15 | 2022-10-25 | 北京踏歌智行科技有限公司 | Mining area environment sensing method based on 4D millimeter wave radar |
CN115656962A (en) * | 2022-12-26 | 2023-01-31 | 南京楚航科技有限公司 | Method for identifying height-limited object based on millimeter wave radar |
CN115825912A (en) * | 2023-01-09 | 2023-03-21 | 南京隼眼电子科技有限公司 | Radar signal processing method and device and storage medium |
CN115825912B (en) * | 2023-01-09 | 2023-05-23 | 南京隼眼电子科技有限公司 | Radar signal processing method, device and storage medium |
CN115840221A (en) * | 2023-02-20 | 2023-03-24 | 上海几何伙伴智能驾驶有限公司 | Method for realizing target feature extraction and multi-target tracking based on 4D millimeter wave radar |
CN116593973A (en) * | 2023-03-29 | 2023-08-15 | 深圳承泰科技有限公司 | Method and system for automatically calibrating installation angle of vehicle-mounted millimeter wave radar |
CN116990773A (en) * | 2023-09-27 | 2023-11-03 | 广州辰创科技发展有限公司 | Low-speed small target detection method and device based on self-adaptive threshold and storage medium |
CN117250595A (en) * | 2023-11-20 | 2023-12-19 | 长沙莫之比智能科技有限公司 | False alarm suppression method for vehicle-mounted millimeter wave radar metal well lid target |
CN117250595B (en) * | 2023-11-20 | 2024-01-12 | 长沙莫之比智能科技有限公司 | False alarm suppression method for vehicle-mounted millimeter wave radar metal well lid target |
CN117491965A (en) * | 2024-01-02 | 2024-02-02 | 上海几何伙伴智能驾驶有限公司 | Target track starting method based on 4D millimeter wave radar |
CN117491965B (en) * | 2024-01-02 | 2024-03-19 | 上海几何伙伴智能驾驶有限公司 | Target track starting method based on 4D millimeter wave radar |
CN117647807A (en) * | 2024-01-30 | 2024-03-05 | 安徽隼波科技有限公司 | Motor vehicle size estimation method based on millimeter wave radar |
CN117647807B (en) * | 2024-01-30 | 2024-04-19 | 安徽隼波科技有限公司 | Motor vehicle size estimation method based on millimeter wave radar |
Also Published As
Publication number | Publication date |
---|---|
DE112022000017T5 (en) | 2023-07-27 |
CN113866742B (en) | 2022-02-22 |
WO2023097971A1 (en) | 2023-06-08 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN113866742B (en) | Method for point cloud processing and target classification of 4D millimeter wave radar | |
US11093799B2 (en) | Rare instance classifiers | |
Wang et al. | A point cloud-based robust road curb detection and tracking method | |
CN109087510B (en) | Traffic monitoring method and device | |
Lin et al. | A Real‐Time Vehicle Counting, Speed Estimation, and Classification System Based on Virtual Detection Zone and YOLO | |
He et al. | Obstacle detection of rail transit based on deep learning | |
CN106228125B (en) | Method for detecting lane lines based on integrated study cascade classifier | |
CN111340855A (en) | Road moving target detection method based on track prediction | |
CN110940971B (en) | Radar target point trace recording method and device and storage medium | |
CN114488194A (en) | Method for detecting and identifying targets under structured road of intelligent driving vehicle | |
CN112990004A (en) | Black smoke vehicle detection method based on optical flow method and deep learning convolutional neural network | |
CN114358140A (en) | Rapid capturing method for sparse point cloud aircraft under low visibility | |
CN114879192A (en) | Decision tree vehicle type classification method based on road side millimeter wave radar and electronic equipment | |
CN117075097B (en) | Maritime radar target tracking method and system based on expanded target cluster division | |
CN117689995A (en) | Unknown spacecraft level detection method based on monocular image | |
Piroli et al. | Towards robust 3D object detection in rainy conditions | |
Huang et al. | An improved YOLOv3‐tiny algorithm for vehicle detection in natural scenes | |
Yang et al. | Learn to model and filter point cloud noise for a near-infrared ToF LiDAR in adverse weather | |
CN113313008B (en) | Target and identification tracking method based on YOLOv3 network and mean shift | |
Qi et al. | Vehicle detection under unmanned aerial vehicle based on improved YOLOv3 | |
CN114895274A (en) | Guardrail identification method | |
CN114170196A (en) | SAR image small target identification method based on CenterNet2 | |
Wang et al. | An Improved Object Detection Method for Underwater Sonar Image Based on PP‐YOLOv2 | |
Abdalwohab et al. | Deep learning based camera and radar fusion for object detection and classification | |
Ding et al. | Lane line detection based on YOLOv4 |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
PB01 | Publication | ||
PB01 | Publication | ||
SE01 | Entry into force of request for substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
GR01 | Patent grant | ||
GR01 | Patent grant |