Target detection method and device, integrated circuit, electromagnetic wave sensor and terminal equipment

Abstract

The invention discloses a target detection method and device, an integrated circuit, an electromagnetic wave sensor and terminal equipment, original point cloud in the target detection method and device can comprise information reflecting physiological signals, and the target detection method and device are suitable for target detection in a short-distance sealed or relatively sealed space area; moreover, according to the embodiment of the invention, the point cloud data is divided into different clusters through clustering based on the original point cloud, and the track of the target is updated and managed based on association between the centroid of each cluster and the track information of the previous frame, so that false alarms or multipath points accidentally appearing in the time sequence in the point cloud result are removed, and the accuracy of the target track is improved. The real target points which disappear accidentally in the time sequence in the point cloud result are compensated, the point cloud quality is improved, the detection rate is effectively improved, the false alarm target number is reduced, and a guarantee is provided for high-precision target detection.

Description

Technical Field

[0001] The present application relates to but is not limited to radar signal processing technology, and in particular to a target detection method and device, an integrated circuit, an electromagnetic wave sensor and a terminal device. Background Art

[0002] When performing target detection in a short-range sealed or relatively sealed space area, such as target detection in a car cabin, there are technical problems such as low detection rate and a large number of false alarm targets due to the influence of strong static clutter and multipath. Summary of the invention

[0003] The present application provides a method and device for detecting a mine target, an integrated circuit, an electromagnetic wave sensor and equipment, which can effectively improve the detection rate, reduce the number of false alarm targets, and provide a guarantee for high-precision target detection.

[0004] An embodiment of the present invention provides a target detection method, comprising: performing cluster analysis on an original point cloud, wherein the original point cloud includes information reflecting physiological signals; tracking a target and generating a target point cloud based on the clustering result and a prediction value of a current frame; and obtaining the prediction value of the current frame based on information of previous frames in the track.

[0005] In an exemplary embodiment, it also includes: using the target point cloud to count the number of points in a preset area; and detecting the target in the detection area according to the number of times the target appears in the preset area.

[0006] In an exemplary embodiment, the original point cloud is a point cloud obtained after processing based on inter-frame fast Fourier transform FFT.

[0007] In an exemplary embodiment, the physiological signal includes: a breathing signal, a heartbeat signal and / or a pulsation signal, etc.

[0008] In an exemplary embodiment, a density-based noise application spatial clustering DBSCAN algorithm, or a partitioning clustering algorithm k-means, or a density-based algorithm OPTICS, or a hierarchical clustering algorithm Agglomerative, etc., is used, as long as the possible targets in the original point cloud can be grouped into target point clusters.

[0009] In an exemplary embodiment, the original point cloud may include one or any combination of information including distance, azimuth, speed and pitch angle of the target.

[0010] In an exemplary instance, the clustering analysis of the original point cloud includes: screening out points that meet the core object requirements from possible targets in the original point cloud, so that the core object contains at least MinPts points in the ε neighborhood; connecting the core points in each other's ε neighborhood to form a point cluster, adding the points in the ε neighborhood where the core point of the cluster is located to the point cluster to form multiple point clusters, and using the center point of each cluster point cloud as the cluster center point.

[0011] In an exemplary instance, the predicted value of the current frame is obtained based on the information of the previous frame in the track, including: using the trajectory of the previous frame in each track, or the average of multiple previous frames, or the median of the track points of the past M frames or some characteristic value in a specified order.

[0012] In an exemplary embodiment, the trajectory of the previous frame in each track is used as the predicted value of the current frame; the target is tracked and the target point cloud is generated based on the clustering result and the predicted value of the current frame, including: calculating the distance between the predicted value of each track and all the cluster center points of the current frame respectively, adding the cluster center point associated with the current track to the corresponding track according to the calculated distance, and updating the track that has matched the cluster center point; generating the target point cloud based on the original point cloud of a single cluster corresponding to each newly added cluster center point in all tracks.

[0013] In an exemplary instance, the cluster center points associated with the current track are added to the corresponding track based on the calculated distance, including: for cluster center points whose calculated distance is less than a preset threshold, they are considered to match the track and the cluster center points are added to the corresponding track; for cluster center points whose calculated distance is greater than or equal to the preset threshold, they are added to the candidate tracks for track start determination.

[0014] In an exemplary embodiment, the distance is: Euclidean distance, Chebyshev distance, Manhattan distance, etc.

[0015] In an exemplary instance, the updating of the tracks that have matched the cluster center points may include: for the existing tracks, if no matching cluster center points are found in the current frame, increasing the end flag values ​​of these tracks; for the tracks that have not matched the cluster center points, checking whether the values ​​of their end flag bits exceed a preset threshold, and if so, removing these tracks; adding the candidate tracks that meet the MN judgment to the formal tracks, and the MN judgment criterion is: if the candidate tracks have N frames that can be associated with the cluster center points within the past M frames, it means that the judgment is successful.

[0016] In an exemplary instance, generating a target point cloud based on the original point cloud of a single cluster corresponding to each newly added cluster center point in all tracks may include: extracting the newly added cluster center points in all tracks, outputting and recording the original point cloud and track number of the single cluster corresponding to each cluster center point to generate the target point cloud.

[0017] In an exemplary embodiment, the method of using the target point cloud to count the number of points in a preset area includes: dividing the entire detection area into a plurality of the preset areas; and counting the target points detected in each frame according to the preset areas.

[0018] In an exemplary embodiment, the detecting the target in the detection area includes: determining whether there is a target in the detection area.

[0019] In an exemplary embodiment, the method further includes: determining a specific area where the target existing in the detection area is located.

[0020] In an exemplary embodiment, before the cluster analysis of the original point cloud is performed, it also includes: filtering out the first DC component of the analog-to-digital conversion ADC data; performing windowing processing on the data after filtering out the first DC component, and obtaining and storing distance dimension data through a one-dimensional fast Fourier transform; returning to the step of filtering out the first DC component on the ADC data until the accumulated distance dimension data reaches a preset number of frames; filtering out the second DC component of the distance dimension data of the preset number of frames; performing windowing processing on the data after filtering out the second DC component, and obtaining distance dimension Doppler data through a two-dimensional fast Fourier transform; performing incoherent accumulation based on the obtained distance dimension Doppler data, and estimating the noise floor of each distance unit of the data after the incoherent accumulation, and performing constant false alarm rate CFAR processing; determining the position and motion state of the possible target in space based on the CFAR detection points obtained after the CFAR processing to obtain the original point cloud.

[0021] In an exemplary embodiment, the preset number of frames is 256 frames, 128 frames, 56 frames, or 32 frames, etc., which can be set based on actual applications.

[0022] An embodiment of the present application further provides a computer-readable storage medium storing computer-executable instructions, wherein the computer-executable instructions are used to execute any of the target detection methods described above.

[0023] An embodiment of the present application further provides an electromagnetic wave sensor, comprising: a signal transmitting module, configured to transmit electromagnetic waves for target detection; a signal receiving module, configured to receive echoes formed by reflection and / or scattering of the electromagnetic waves; and a processing module, configured to perform signal and data processing on the echoes to achieve operations such as determination, positioning and / or identification of the target; wherein the processing module is also configured to implement any target detection method provided in the embodiments of the present application.

[0024] In an exemplary embodiment, when the method includes digital beamforming and frame data accumulation, the baseband unit may be configured to implement the digital beamforming, and the MCU unit may be configured to implement the frame data accumulation.

[0025] An embodiment of the present application further provides an integrated circuit, comprising a radio frequency module, an analog signal processing module and a digital signal processing module connected in sequence; the radio frequency module is used to generate an electromagnetic wave signal and receive an echo signal; the analog signal processing module is used to down-convert the echo signal to obtain an intermediate frequency signal; the digital processing module is used to perform analog-to-digital conversion on the intermediate frequency signal to obtain a digital signal; and the digital processing module performs target judgment, positioning and / or identification operations based on any of the target detection methods described above.

[0026] In an exemplary embodiment, the integrated circuit is a millimeter wave chip, or a sensor chip.

[0027] This embodiment also provides an electromagnetic wave sensor, including: a carrier; the above-mentioned integrated circuit, which is arranged on the carrier; an antenna, which is arranged on the carrier, or the antenna and the integrated circuit are integrated into a single device and arranged on the carrier; wherein the integrated circuit is connected to the antenna for transmitting the electromagnetic wave signal and / or receiving the echo signal.

[0028] An embodiment of the present application further provides a terminal device, comprising: a device body; and the above-mentioned electromagnetic wave sensor arranged on the device body; wherein the electromagnetic wave sensor is used for target detection and / or communication to provide reference information for the operation of the device body.

[0029] The radar data processing method provided in the embodiment of the present application uses a multi-frame joint processing technology, adopts a new inter-frame accumulation method, and combines the detected point cloud clustering and track processing methods to effectively filter out false alarm points and multipath target points, thereby improving the quality of the detected point cloud, thereby providing a guarantee for improving the detection, measurement or recognition performance, and further effectively improving the detection rate, reducing the number of false alarm targets, and thus improving the accuracy of target detection.

[0030] Furthermore, the radar data processing method provided in the embodiment of the present application, while effectively improving the detection rate and reducing the number of false alarm targets, also greatly improves the accuracy of angle estimation and other effects, thereby enabling special targets or weak targets such as infants and young children in the cabin to be accurately detected, thereby realizing applications such as CPD and SBR.

[0031] In some embodiments, the target detection method provided in the embodiments of the present application is used to achieve target detection and recognition in spatial areas such as car cabins, indoors, and factory buildings, and in particular has a better detection effect on achieving in-car detection of babies (targets).

[0032] Other features and advantages of the present invention will be described in the following description, and partly become apparent from the description, or understood by practicing the present invention. The purpose and other advantages of the present invention can be realized and obtained by the structures particularly pointed out in the description, claims and drawings. BRIEF DESCRIPTION OF THE DRAWINGS

[0033] The accompanying drawings are used to provide further understanding of the technical solution of the present application and constitute a part of the specification. Together with the embodiments of the present application, they are used to explain the technical solution of the present application and do not constitute a limitation on the technical solution of the present application.

[0034] Figure 1 Schematic diagram of the process of target detection method in the embodiment of the present application;

[0035] Figure 2 This is a schematic diagram of clustering by density using the DBSCAN algorithm in an embodiment of the present application;

[0036] Figure 3 A schematic diagram of detecting a target using a target point cloud within a detection area in an embodiment of the present application;

[0037] Figure 4 This is a flow chart of realizing target detection based on SISO-combine in an embodiment of the present application;

[0038] Figure 5 This is a flow chart of implementing target detection based on per-channel in an embodiment of the present application;

[0039] Figure 6 It is a flowchart of realizing target detection based on frame-level FFT combined with DAE CFAR in an embodiment of the present application;

[0040] Figure 7 Based on Figure 6 A schematic diagram of azimuth or elevation CFAR in the target detection process shown;

[0041] Figure 8This is a schematic diagram of a process for post-processing a target in an embodiment of the present application;

[0042] Fig. 9 This is a flow chart of post-processing a target in combination with a DL algorithm in an embodiment of the present application;

[0043] Fig.10 This is a flow chart of a target detection method combined with a DL algorithm in an embodiment of the present application;

[0044] Fig.11 It is a flowchart of another target detection method combined with DL algorithm in an embodiment of the present application;

[0045] Fig.12 Schematic diagram of the flow of another target detection method combined with DL algorithm in an embodiment of the present application.

[0046] Fig.13 This is a schematic diagram of a flow chart of a radar data processing embodiment in a target detection method in an embodiment of the present application;

[0047] FIG14( a ) is a schematic diagram of a point cloud result in the related art under the first test scenario;

[0048] FIG14( b ) is a schematic diagram of a point cloud result after processing by an embodiment of the present application in the first test scenario;

[0049] FIG15( a ) is a schematic diagram of a point cloud result in the related art under the second test scenario;

[0050] FIG15( b ) is a schematic diagram of a point cloud result after processing by an embodiment of the present application in a second test scenario;

[0051] FIG16( a ) is a schematic diagram of a point cloud result in the related art under the third test scenario;

[0052] FIG16( b ) is a schematic diagram of a point cloud result after processing by an embodiment of the present application in a third test scenario;

[0053] FIG17( a ) is a schematic diagram of a point cloud result in the related art under the fourth test scenario;

[0054] FIG17( b ) is a schematic diagram of a point cloud result after processing by an embodiment of the present application in the fourth test scenario;

[0055] Fig.18 This is a schematic diagram of the structure of the target detection device in the embodiment of the present application;

[0056] Fig.19 A schematic diagram of the structure of an object detection device in an embodiment of the present application;

[0057] Fig. 20It is a schematic diagram of the composition structure of another embodiment of the target detection device in the embodiments of the present application. DETAILED DESCRIPTION

[0058] In order to make the purpose, technical solution and advantages of the present application more clear, the embodiments of the present application will be described in detail with reference to the accompanying drawings. It should be noted that the embodiments and features in the embodiments of the present application can be combined with each other arbitrarily without conflict.

[0059] In order to facilitate understanding of the present application, the present application will be described more fully below with reference to the relevant drawings. Embodiments of the present application are provided in the drawings. However, the present application can be implemented in many different forms and is not limited to the embodiments described herein. On the contrary, the purpose of providing these embodiments is to make the disclosure of the present application more thorough and comprehensive.

[0060] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as those commonly understood by those skilled in the art to which this application belongs. The terms used herein in the specification of this application are only for the purpose of describing specific embodiments and are not intended to limit this application.

[0061] It is understood that the terms "first" and "second" used in this application are only used for descriptive purposes and cannot be understood as indicating or implying relative importance or implicitly indicating the number of technical features indicated. Therefore, the features defined as "first" and "second" may explicitly or implicitly include at least one of the features. In the description of this application, the meaning of "plurality" is at least two, such as two, three, etc., unless otherwise clearly and specifically defined.

[0062] It can be understood that the “connection” in the following embodiments should be understood as “electrical connection”, “communication connection”, etc. if the connected circuits, modules, units, etc. have electrical signals or data transmission between each other.

[0063] When used herein, the singular forms "a", "an", and "said / the" may also include plural forms, unless the context clearly indicates otherwise. It should also be understood that the terms "include / comprise" or "have" and the like specify the presence of stated features, wholes, steps, operations, components, parts, or combinations thereof, but do not exclude the possibility of the presence or addition of one or more other features, wholes, steps, operations, components, parts, or combinations thereof. At the same time, the term "and / or" used in this specification includes any and all combinations of the relevant listed items.

[0064] The steps shown in the flowchart of the accompanying drawings can be executed in a computer system such as a set of computer executable instructions. Also, although a logical sequence is shown in the flowchart, in some cases, the steps shown or described can be performed in a sequence different from that shown here.

[0065] After the radar system obtains the original point cloud including information such as the distance, azimuth, pitch angle, speed, etc. of the target, how to perform regional judgment based on the obtained original point cloud to achieve target detection is crucial to effectively filter out unstable false alarms and multipath targets, improve the quality of point clouds, and thus improve detection and recognition performance. For the implementation method of performing Capon processing based on single-frame one-dimensional FFT data to generate RA spectrum, performing CFAR detection to generate point cloud, and then performing regional judgment, since the single-frame information is greatly affected by static clutter, the point cloud quality is low, which is prone to false alarms and missed detections. For the implementation method of directly executing regional statistics and judgment algorithms on the original point cloud, the statistical results are smoothed by increasing the cache time of the statistical results. Although reasonable thresholds and regional mutual exclusion rules are added to the regional judgment algorithm to achieve accurate positioning of the personnel in the cabin, in some cases such as when there are noise points and there are fewer valid point clouds, or when there are many noise points and there are fewer valid point clouds, this implementation method may lead to regional judgment errors.

[0066] In order to effectively filter out unstable false alarms and multipath targets, improve the quality of point clouds, and further improve detection and recognition performance, an embodiment of the present application provides a target detection method, which aims to use a time series association method to perform cluster analysis, track processing, regional history statistics and regional decision-making on the detected original point cloud, and ultimately achieve regional determination to accurately locate the target position. Figure 1 Schematic diagram of the process of the target detection method in the embodiment of the present application, such as Figure 1 As shown, including:

[0067] Step 100: performing cluster analysis on the original point cloud, where the original point cloud includes information reflecting physiological signals.

[0068] In an exemplary embodiment, the original point cloud may be a point cloud obtained after determining the position and motion state of a possible target in space based on CFAR detection points obtained after CFAR processing.

[0069] The original point cloud can be represented by four-dimensional information (x, y, z, v) of possible targets in a Cartesian coordinate system, where x can represent the horizontal direction, y can represent the vertical direction, and z can represent the longitudinal direction.

[0070] The Cartesian coordinate system (x, y, z) corresponds to the polar coordinate system (target distance r, azimuth angle θ, pitch angle φ), and the velocity v is usually obtained directly from the target's Doppler information and remains unchanged.

[0071] In an exemplary embodiment, physiological signals refer to various electrical, magnetic, optical, acoustic, force and other signals generated by the life activities of an organism, and physiological signals can reflect the physiological state and functional activities of the organism. In an embodiment, physiological signals may include, but are not limited to, respiratory signals and / or heartbeat signals.

[0072] Taking the breathing signal as an example, breathing causes the expansion and contraction of the chest cavity, which will cause slight changes in distance. These slight changes in distance can be reflected in the point cloud data of the distance dimension. During breathing, the movement of the chest cavity will cause slight changes in speed. These slight changes in speed can be reflected in the point cloud data of the Doppler dimension. These changes can be identified by extracting the speed information. Although the movement caused by breathing is mainly changes in the distance and speed dimensions, there may also be slight changes in azimuth and pitch angles, especially when the body moves slightly. These slight changes can be reflected in the point cloud data of the azimuth and pitch angle dimensions. Therefore, physiological signals can be reflected by the data such as distance, speed, azimuth and pitch angle included in the original point cloud.

[0073] In the embodiment of the present application, the original point cloud includes at least one or any combination of the distance, azimuth and pitch angle of the target. Further, the original point cloud may also include speed, where the speed information can characterize the law of the target's reciprocating motion (such as jumping), such as period, frequency, change trend, etc.

[0074] In an exemplary embodiment, the original point cloud may be clustered using a density-based spatial clustering of applications with noise (DBSCAN) algorithm using x, y, z information in the four-dimensional information (x, y, z, v) of the possible target to group the CRAF detection points (i.e., possible targets) into target point clusters. In an embodiment, the clustering analysis of the original point cloud in step 100 may include:

[0075] Filter out points that meet the requirements of core objects from the possible targets in the original point cloud, that is, core objects that contain at least MinPts points in the ε neighborhood;

[0076] The core points in each other's ε neighborhood are connected to form a point cluster, and the points in the ε neighborhood of the core point of the cluster are also added to the point cluster, eventually forming multiple point clusters, and the center point of each cluster point cloud is used as the cluster center point.

[0077] In some embodiments, the center point of each cluster point cloud can be the average value (center of mass) of each cluster point cloud, or the median value of each cluster point cloud can be taken as the cluster center point, which can better reduce the impact of large deviations in each cluster point cloud.

[0078] Epsilon Neighborhood is used to define the density connection relationship between points. Specifically, the ε neighborhood refers to the set of all points within a circle (in two-dimensional space) or a sphere (in three-dimensional space) with a radius of ε around a specified point; Figure 2 This reflects how the DBSCAN algorithm performs clustering by density, such as Figure 2 As shown in the figure, black dots represent data points, which can be CRAF detection points used to represent possible targets; black circles represent the ε neighborhood of each point, that is, the area with a radius of ε centered on each point; gray dots represent core points, which are cores that contain at least MinPts points in their ε neighborhood. Arrow lines represent the connection between core points, indicating that these core points are connected to each other to form a point cluster. Among them, the number of MinPts points can be determined according to the actual application scenario. For example, if the goal is to identify smaller or sparser clusters, a smaller MinPts can be selected. For example, if the goal is to identify larger or denser clusters, a larger MinPts can be selected.

[0079] In an exemplary embodiment, the clustering algorithm may also be a partitioning clustering algorithm k-means, a density-based algorithm OPTICS, a hierarchical clustering algorithm Agglomerative, and the like.

[0080] In some embodiments, the clustered original data may also be, but is not limited to, various combinations such as: four-dimensional information (x, y, z, v) of possible targets, x, y information in the four-dimensional information (x, y, z, v) of possible targets, x, z information in the four-dimensional information (x, y, z, v) of possible targets, y, z information in the four-dimensional information (x, y, z, v) of possible targets, and x, y, v information in the four-dimensional information (x, y, z, v) of possible targets.

[0081] Step 101: According to the clustering results and the prediction value of the current frame, the target is tracked and a target point cloud is generated; the prediction value of the current frame is obtained based on the information of the previous frame in the track.

[0082] In an exemplary instance, the predicted value of the current frame can be obtained based on the information of the previous frame in the track, which may include but is not limited to: using the trajectory of the previous frame in each track, or the average of the track points of multiple previous frames such as M frames, or the median of the track points of the past M frames, or some characteristic value in a specified order (or sorting), etc.

[0083] In an exemplary instance, step 101 may include: taking the trajectory of the previous frame in each track as the predicted value of the current frame; calculating the distance between the predicted value of each track and all cluster center points of the current frame, adding the cluster center point associated with the current track to the corresponding track according to the calculated distance, and updating the track that has matched the cluster center point; generating a target point cloud based on the original point cloud of a single cluster corresponding to each newly added cluster center point in all tracks.

[0084] In an exemplary embodiment, a track is a continuous position record of a target over multiple time frames. Each track contains the position information of the target at each time point. For example, a track can be represented as a set of position points: [pos1, pos2, ...pos n ], where pos i Indicates the position of the target in the i-th frame. Generally speaking, the technology of track processing is mostly used for track tracking and management of high-speed moving targets. The movement of the target causes its spatial position to change. Usually, methods such as Kalman filtering, extended Kalman filtering, particle filtering, etc. can be used to predict the four-dimensional information (x, y, z, v) of the next frame track based on the four-dimensional information (x, y, z, v) in the previous frame track. However, the embodiments of the present application are mainly aimed at short-distance sealed or relatively sealed space areas, and the application scenarios where the target spatial position changes with a small amplitude and is independent of speed. Therefore, the track information of the previous frame is used as the predicted value in the embodiments of the present application. In this way, the calculation of coefficients such as the state space equation and the error covariance matrix is ​​avoided, and the calculation time is reduced.

[0085] It should be noted that the method for selecting the prediction value may include but is not limited to: using the mean, median or certain characteristic value of the track points of the past M frames in a specified order (or sorting) as the prediction value of the current frame.

[0086] In one embodiment, in a CPD scenario, the position of the CPD target point is relatively stable. Unlike ADAS applications, the target Doppler information in the CPD only represents the micro-motion frequency characteristics (such as breathing and heart rate), which does not affect the change of the target spatial position. Therefore, the trajectory prediction method can be simplified. In an embodiment of the present application, the trajectory point of the previous frame in each track is used as the prediction value of the current frame, without the need for complex filtering methods such as Kalman filtering, thereby saving computational costs.

[0087] In an exemplary embodiment, cluster center points associated with the current track are added to the corresponding track based on the calculated distance, which may include: cluster center points whose calculated distance is less than a preset threshold are considered to match the track, and the cluster center points are added to the corresponding track, which means that these cluster center points are considered to be associated with the track, and may be the new position of the track in the current frame or the continued trajectory; for unmatched cluster center points, that is, cluster center points whose calculated distance is greater than or equal to the preset threshold, they are added to the candidate tracks for track start determination, that is, these candidate tracks may be newly appeared targets or the starting part of the original track, and further observation and analysis are required to determine whether they form a valid track.

[0088] In an exemplary embodiment, the distance calculation method may include but is not limited to: based on the nearest neighbor (NN) method, the Euclidean distance D between the predicted value of each track and all cluster centers of the current frame is calculated as: This formula is used to determine the match between the predicted value of the track and the cluster center point; or, Chebyshev distance D = max(|x1-x2|,|y1-y2|,|z1-z2|); or, Manhattan distance (L1 distance) D = |x1-x2|+|y1-y2|+|z1-z2|.

[0089] In an exemplary embodiment, the Euclidean distance between the track prediction value and each cluster center point is calculated based on the NN algorithm to represent the correlation between the center point and the track. In this embodiment, only the x, y, z information in the four-dimensional information (x, y, z, v) is used. In some embodiments, the v information may also be added, that is, the four-dimensional information is used for association. In other embodiments, for radars without pitch angle measurement capabilities, only the x, y, v information in the four-dimensional information (x, y, z, v) or the x, y information in the four-dimensional information (x, y, z, v) may be used for association.

[0090] In an exemplary embodiment, updating the tracks that have matched the cluster center point may include: for the existing tracks, if no matching cluster center point is found in the current frame, then the end flag value of these tracks will be increased. This end flag can be used to determine whether the track has ended or temporarily interrupted for subsequent track management and analysis. For tracks that have not matched the cluster center point, check whether the value of their end flag exceeds the preset threshold. If so, remove these tracks; add candidate tracks that meet the MN judgment to the formal track. The MN judgment criterion is: if the candidate track has N frames that can be associated with the cluster center point within the past M frames, it means that the judgment is successful.

[0091] In an exemplary instance, generating a target point cloud based on the original point cloud of a single cluster corresponding to each newly added cluster center point in all tracks can include: extracting the newly added cluster center points in all tracks, outputting and recording the original point cloud of a single cluster corresponding to each cluster center point and the track number to generate the target point cloud.

[0092] In step 100 and step 101 of the embodiment of the present application, point cloud data is divided into different clusters by clustering, and the centroid (center point) of each cluster is calculated, and then the target track is updated and managed based on the association of these centroids with the trajectory of the previous frame to achieve trajectory prediction and matching.

[0093] In the embodiment of the present application, the trajectory prediction of step 101 is used to associate and allocate the track, and the isolated target points are removed. In other words, through the point cloud track processing of step 101, the false alarms or multipath points that occasionally appear in the time sequence in the point cloud results are removed, and the real target points that occasionally disappear in the time sequence in the point cloud results are compensated, thereby improving the point cloud quality, effectively improving the detection rate, and reducing the number of false alarm targets, providing a guarantee for high-precision target detection.

[0094] In an exemplary embodiment, the target detection method provided by the embodiment of the present application further includes: Step 102: using the target point cloud to count the number of points in a preset area.

[0095] In an exemplary embodiment, step 102 may include: dividing the entire detection area into a plurality of preset areas; and counting the target points detected in each frame according to the preset areas respectively.

[0096] In one embodiment, for the CPD scenario, the detection area is all spaces in the vehicle, including all seats, aisles, and other areas where infants or people may be present. The preset area may be, for example, a seat space area, an aisle space area, and the like.

[0097] Step 103: Detect the target in the detection area according to the number of times the target appears in the preset area.

[0098] In an exemplary embodiment, detecting a target in a detection area may include: determining whether a target exists in the detection area. Further, it may include: determining a specific area where the target exists in the detection area.

[0099] In one embodiment, for a CPD scenario, the preset area may be an area of ​​interest such as a seat space, an aisle space, etc.

[0100] In an exemplary embodiment, taking the detection of people in a vehicle as an example, in counting the target points detected in each frame according to the preset areas, an implementation method of the preset areas can be as follows: Figure 3 As shown, Figure 3The horizontal coordinates are the X direction (vehicle width direction) and the Z direction (vehicle tail-to-vehicle front direction). The interior space of the vehicle is divided into areas that need to be determined along all or some of the XYZ dimensions, or along all or some of the distance-azimuth-pitch dimensions of the polar coordinate system. Figure 3 The vehicle is divided into six areas: three seats in the back row (seat A, seat B, seat C) and three aisles in front of the three seats (aisle A, aisle B, aisle C). These areas may overlap with each other or not; some areas in the vehicle may belong to multiple preset areas at the same time or not belong to any area. In one embodiment, the counting steps are performed for the target points detected in each frame according to the preset areas. Assuming that 11 target points are detected in the current frame in the target point cloud in this embodiment, the number of these target points in the above preset areas is counted, and the distribution of the 11 target points is obtained as follows: seat A (6), seat B (2), seat C (0), aisle A (0), aisle B (1), and aisle C (0). In one embodiment, for the step of determining whether there is a target in the detection area and the specific area where it is located, first, set the flag bit and judgment condition: initialize the flag bit flag_region_i of each preset area used to determine whether there is a person to be 0, that is, no one is present (i=0,1,...,5); if the total number of valid targets total_valid_tgt_num is 0, directly determine that there is no one in the car and end; traverse all preset areas, if the number of target points tgt_num_region_i of a preset area i exceeds 25% of the total number of valid targets total_valid_tgt_num, set the flag bit flag_region_i of the preset area i to 1. Next, determine whether there is someone in the car: if the flag bit flag_region_i of a preset area is 1, set flag_region_empty to 0, indicating that there is someone in the car. After that, the seats and aisles are judged: the seat areas are traversed, and if the flag flag_region_i of a seat area is 1, it is judged that the corresponding aisle area is empty. For example, if the flag of seat A is 1, which means there is someone, the flag of aisle A is set to 0, which means there is no one. Finally, according to the setting of the flag, the final judgment result is output, including whether there is someone and the specific area where they are located.

[0101] Combination Figure 3, take 11 target points detected, and their distribution in each preset area is: seat A: 6; seat B: 2; seat C: 0; aisle A: 0; aisle B: 1; aisle C: 0 as an example, the total valid target number total_valid_tgt_num = 11, then, traversing the area, we can get: seat A: 6 (accounting for 54.5% of the total> 25%), set flag_region_0 = 1; seat B: 2 (accounting for 18.2% of the total< 25%). %), keep flag_region_1=0; Seat C: 0 (0%<25% of the total), keep flag_region_2=0; Aisle A: 0 (0%<25% of the total), keep flag_region_3=0; Aisle B: 1 (9.1%<25% of the total), keep flag_region_4=0; Aisle C: 0 (0%<25% of the total), keep flag_region_5=0. Since flag_region_0=1 exists, it is determined that there is someone in the car, and flag_region_empty=0 is set. Moreover, since flag_region_0=1 (there is someone in seat A), flag_region_3=0 is set (there is no one in aisle A). Therefore, the output result is: there is someone in the car, and there is someone in seat A, no one in aisle A, and no one in other areas.

[0102] The target detection method provided in the embodiment of the present application includes information reflecting physiological signals in the original point cloud, and is used for target detection in short-distance sealed or relatively sealed spatial areas; moreover, in the embodiment of the present application, the point cloud data is divided into different clusters through clustering based on the original point cloud, and the target track is updated and managed based on the association between the centroid of each cluster and the trajectory information of the previous frame, thereby removing false alarms or multipath points that occasionally appear in the point cloud results in time sequence, compensating for real target points that occasionally disappear in time sequence in the point cloud results, improving the point cloud quality, effectively improving the detection rate, and reducing the number of false alarm targets, providing a guarantee for high-precision target detection.

[0103] The present application also provides a computer-readable storage medium storing computer-executable instructions, wherein the computer-executable instructions are used to execute any of the target detection methods described above.

[0104] The present application further provides a computer device, comprising a memory and a processor, wherein the memory stores the following instructions executable by the processor: for executing the steps of any of the target detection methods described above.

[0105] When performing target detection in a sealed or relatively sealed space area, in some optional embodiments, target detection can be achieved by performing regional judgment based on single-frame fast and slow time processing, constant false alarm rate (CFAR) detection, angle measurement obtained point cloud, etc., or by using Capon and other algorithms to perform operations such as arrival angle (DoA) (i.e., wave direction estimation) based on single-frame fast time processing, thereby forming a range-azimuth (RA, Range-Azimuth) diagram, extracting features such as energy in a predetermined area, and performing identification and judgment to achieve target detection, or by detecting human breathing and heartbeat after filtering the phase of the received echo. For example, when target detection is achieved by performing regional judgment based on single-frame fast and slow time processing, CFAR detection, angle measurement obtained point cloud, after the echo signal is processed such as distance dimension fast Fourier transform (1D-FFT), time correlation analysis can be performed based on complex echo data to achieve human target discrimination, and then position judgment can be performed by combining point cloud information, etc., to ultimately achieve the purpose of cabin target detection.

[0106] In some optional embodiments of the present application, another target detection and / or recognition scheme is proposed, that is, a scheme for realizing target detection in sealed space areas such as cabins, indoors, and factories through inter-frame accumulation, so as to effectively improve the detection rate, reduce the number of false alarm targets, and greatly improve the accuracy of angle estimation, etc., so that special targets or weak targets such as infants and young children in the cabin can be accurately detected, realizing applications such as child presence detection (CPD) and safety belt reminder (SBR).

[0107] In the specific implementation process, accurate detection of targets in the cabin can be achieved based on a combination of one or at least two solutions including multi-frame joint processing technology, deep learning detection positioning and recognition technology, etc.

[0108] The multi-frame joint processing technology can be to obtain the range-Doppler map by sliding window FFT on the multi-frame data, or to use finite impulse response (FIR) or other complex time-frequency transformation processing, and then perform CFAR, DOA and other processing operations on the area of ​​interest, and can be combined with the set area judgment logic and area parameters to achieve the detection and positioning of living targets such as adults, children, pets, or other non-living targets. CFAR can be a non-coherent constant false alarm rate (NR-CFAR, Non-Coherent CFAR) in the Doppler dimension, or a range-Doppler constant false alarm rate (RD-CFAR, Range-Doppler CFAR), or a Doppler-azimuth-elevation constant false alarm rate (DAE-CFA, Doppler-Azimuth-Elevation CFAR), etc. At the same time, after CFAR and DoA processing, post-processing such as clustering, false alarm suppression, and multiple processing point cloud association can be used. The regional parameters can be determined by at least one of the following schemes or a combination of at least two of the following schemes: by clustering the point cloud detected after a certain sliding window multi-frame processing, performing outlier detection and elimination operations on the point cloud detected after a certain sliding window multi-frame processing, etc.

[0109] Among them, when applied to a closed space area or a relatively closed space area, the space area can be pre-divided to achieve the definition and detection of different areas of interest (divided areas). For example, when detecting targets in the cabin of a car, the radar monitored area can be divided into seat intervals and aisle intervals, etc., and then corresponding parameter types and thresholds can be preset for different types of areas, and the corresponding processing steps can be combined to achieve accurate detection of targets of interest in specific areas.

[0110] In some optional embodiments, for family cars, the space area in the car can generally be simply divided into head space, rear space and luggage space. If the rear space is a key monitoring area, the rear space can be further divided into seat intervals and aisle intervals, etc. At the same time, the seat interval can also be divided into a corresponding number of seat interval units based on the seats. Similarly, the aisle interval can also be divided into a corresponding number of aisle interval units corresponding to the above-mentioned seat interval units. For example, for a five-seat family car, for the three seats in the rear space, three seat interval units can be correspondingly divided, and three aisle interval units in front of the corresponding three seats. At the same time, for different types of areas (or intervals or interval units), corresponding parameter types and thresholds can be preset, and adaptive signal data processing methods and steps can be used to achieve accurate detection of the target of interest (specific target) in a specific interval or interval unit. Among them, adjacent interval units can have a partial overlapping area, or they can be adjacent or separated by a gap of a preset width.

[0111] The detection, positioning and identification scheme of deep learning can be: after performing 1D-FFT processing on the echo signal, continue with multi-frame sliding window FFT, and perform N-point 2D-DBF processing to form an N-channel range-Doppler (RD, Range-Doppler) spectrum, and then use the deep learning model to perform monitoring, positioning and identification operations based on the N-channel RD spectrum; or, after performing 1D-FFT processing on the echo signal, continue with N-point 2D-DBF processing, and then use the deep learning model to perform monitoring, positioning and identification operations based on the N-channel RD spectrum; or, first perform 1D-FFT processing on the echo signal, then continue with N-point 2D-DBF processing, and then continue with multi-frame sliding window FFT to form an N-channel RD spectrum, and then use the deep learning model to perform monitoring, positioning and identification operations based on the N-channel RD spectrum. Among them, the above-mentioned N can be the number of interval units. Taking the back row interval with three seats in the back row and dividing it into three seat interval units and corresponding three aisle interval units as an example, then N can be equal to 6 at this time.

[0112] Based on the above implementation scheme, after testing with real data, the top-mounted radar installation scenario can effectively achieve a high detection rate, as well as extremely low missed detection rate and false alarm rate. For example, when detecting targets in the cabin, it can achieve a detection rate of more than 99%, a missed detection rate of less than 0.5%, and a false alarm rate of less than 1%.

[0113] The following is a detailed description of the technical solution of the present application based on the technical idea of ​​frame-level Fourier transform with reference to the accompanying drawings, that is, after performing Fourier transform of the chirp in the echo signal in the fast time dimension, accumulating at least two frames of Fourier transform data in the fast time dimension, performing frame-dimensional Fourier transform of the at least two frames of data, and obtaining a range-Doppler (Range-Doppler, RD for short) spectrum, so as to estimate the target distance and / or speed based on the RD spectrum.

[0114] Figure 4 FIG. 1 is a flow chart of realizing target detection based on single-input single-output signal merging (SISO-combine) in an embodiment of the present application. Figure 4 As shown, after the echo signal is subjected to operations such as ADC (analog-to-digital) conversion and sampling processing, range DC removal (Range DC removal) and range Fourier transform (Range FFT, which can also be expressed as 1D-FFT or fast time FFT) are sequentially performed, and then multi-frame accumulation (such as storing 128 frames of data, store 128 frames) is performed, and Doppler DC removal (Doppler DC removal), frame Fourier transform (Frame FFT), non-coherent integration (Non-coherent integration) are continued, and then constant false alarm detection (such as peak-based CFAR (detection with peak selection)) is performed based on the noise obtained by noise estimation (such as using a noise estimator based on variance); finally, angle detection and target classification are performed to obtain the distance, speed and / or angle information of the target; for example, azimuth estimation and elevation estimation can be first performed based on the CFAR result, which can be specifically performed through DBF (digital beam It is implemented by technologies such as digital beamforming (forming) and DoA (direction of arrival estimation).

[0115] In some optional embodiments, for an electromagnetic wave sensor having a BB (Baseband) unit and an MCU (Microcontroller Unit) unit, it can be based on Figure 4As shown in the flowchart of target detection, frame data accumulation (store 128 frames) and target classification (Region HIST, classification) can be performed in the MCU module, while range DC removal (Range DC removal) and Doppler DC removal (Doppler DC removal) can be performed in the BB module or the MCU module. Figure 4 In the process shown, DC filtering of downlink ADC data in the BB module is taken as an example for explanation.

[0116] like Figure 4 As shown, after performing 1D-FFT processing in the BB module or obtaining distance dimension information, the 1D-FFT data can be cached in the MCU module, and when the preset data amount (such as 128, 56, or 32 frames) is cached, the data cached to the preset number of frames can be DC removed and frame-level FFT (frame-level FFT) can be performed in the frame dimension to obtain an RD (Range-Doppler) diagram.

[0117] When the system contains multiple channels, such as Figure 4 As shown, non-coherent integration can also be performed on the RD spectra of multiple channels, and the noise floor of each range bin in the accumulated data can be estimated using the NVE module. For example, the noise floor estimation can be performed based on a preset formula, and subsequent NR (Noise Reference)-CFAR processing can be performed based on the value of the noise floor estimation. The above preset formula can be: ni=min(ni, α*ng), ni is the noise floor estimation of the i-th range bin, ng is the noise floor estimation of the last range bin of interest, and α is a coefficient, which can be set and updated based on demand or engineering data, experience, etc.

[0118] For angle estimation, see Figure 4 As shown, the azimuth dimension DBF and DoA, as well as the elevation dimension DBF and DoA, can be performed on the CFAR detection point data to obtain the azimuth and elevation angle estimation data of each CFAR detection point. In addition, the target points detected in each frame (or preset data amount, i.e., each time) can be counted according to the preset (or divided) area unit design, and the counting results can be judged according to the preset rules to obtain whether there is a physical target of interest (such as an adult, child, infant or pet, etc.) in this processing, and determine its location area.

[0119] In the above-mentioned embodiment, the frame-level FFT is used to replace the traditional chirp-level Doppler FFT to determine the speed of the target so as to achieve accurate detection of the target of interest, and the sliding window FFT processing can be performed on multiple frames to further improve the update frequency of the result, thereby improving the real-time performance of the system. At the same time, the amount of data processed can be reduced by processing only the distance and / or Doppler area of ​​interest. In addition, when estimating the background noise, the background noise is obtained by taking the minimum value based on the estimate of the current range bin combined with the estimate of the last range bin of interest, which can be more suitable for application scenarios in relatively sealed environments such as in the cabin of a car. When judging the regional logic, by counting the target points detected in each frame according to the preset area design, and judging the counting results according to the preset rules, it is possible to accurately judge whether there is a target in each preset area.

[0120] Figure 5 FIG. 1 is a flow chart of realizing target detection based on each channel (per-channel) in an embodiment of the present application. Figure 5 As shown, based on Figure 4 Based on the process structure shown in FIG. 1 , a scheme for incoherently accumulating multi-channel RD spectra into single-channel data for CFAR is proposed. In this embodiment, CFAR processing is performed on all channels one by one (i.e. Figure 5 After the per channel Doppler NR-CFAR and peak detection (Per Channel Doppler NR-CFAR&PeakDetect) in the channel domain, the binary integration in channel domain is continued (i.e. Figure 5 The channel domain binary integration (Binary Integration in Channel-Domain) is then used to estimate the horizontal / azimuth and elevation angles. Specifically, for a system with M channels, each channel is subjected to the following Figure 4After the inter-frame FFT shown in , CFAR processing is performed on each channel to obtain M bit masks, each element of which can be of bool type, and each range unit (rangebin) and / or Doppler unit (Dopplerbin) is judged to determine whether there is a target (if it is 1, it is defined as existing, and if it is 0, it is defined as not existing). At the same time, these M bit masks can be accumulated, and the range of each element can be defined as 0 to (M-1). Subsequently, the accumulated mask can be subjected to secondary detection, that is, each element of the accumulated mask can be compared with a preset threshold value M0. If the value xi>M0 of the unit to be detected (such as an element of the accumulated mask), it can be output that there is a target in the unit to be detected. If xi≤M0, it can be output that there is no target in the unit to be detected. Among them, xi, M, and M0 are positive integers.

[0121] exist Figure 5 In the detection method flow shown, since each channel is subjected to CFAR processing separately, the performance loss caused by channel imbalance can be effectively reduced, and the detection method has better robustness in practical applications.

[0122] Figure 6 FIG. 1 is a flow chart of target detection based on frame-level FFT combined with DAE CFAR in an embodiment of the present application. Figure 6 As shown, in Figure 4 Based on the process structure shown, based on the frame-dimensional FFT (frame-FFT), DAE CFAR (Constant False Alarm Rate) is used to estimate the azimuth and elevation angles for CFAR data. Specifically, after Doppler dimension CFAR detection is performed on the inter-frame FFT data, azimuth dimension DBF (Azimuth DBF) is performed, and the azimuth CFAR processing (AzimuthCFAR) is continued in combination with the noise variance estimation (NVE, Noise Variance Estimator) noise floor estimation; similarly, the elevation dimension DBF (Elevation DBF) is performed on the azimuth CFAR data, and the elevation CFAR processing (Elevation CFAR) is continued in combination with the NVE noise floor estimation, thereby obtaining the azimuth and elevation information of the target.

[0123] Figure 7 Based on Figure 6 The schematic diagram of the azimuth or elevation CFAR in the target detection process is shown. Figure 7 The horizontal axis is the azimuth, and the vertical axis is the amplitude, such as the signal strength or signal-to-noise ratio (SNR). Figure 7As shown, the azimuth / pitch energy data (power data) can be calculated by the NVE engine in a HIST manner, and the azimuth / pitch DBF noise floor estimation can be performed by selecting a certain rank value (such as mean, median, etc.). When performing DBF, all peaks in the DBF spectrum can be obtained, and the SNR of all peak points and the estimated noise floor can be used to determine whether to output it as a peak. For example, when performing DBF, the amplitudes of all peak points in the DBF spectrum can be obtained, including the peak of the global maximum (Global max) and the peak of the local maximum (Local max); for any peak, if its SNR exceeds a predetermined threshold, and the difference between its amplitude (local max) and the global maximum amplitude is within a preset range, and the number of peaks that have been output is less than the preset number, then the peak is output, that is, it is output as a real target point.

[0124] Figure 8 Schematic diagram of the process of post-processing the target in the embodiment of the present application. Figure 8 As shown, it can be Figure 4 Based on the process structure shown, after the azimuth / elevation angle estimation, the output target point cloud data can be processed in combination with the machine learning model to accurately detect the target in the area of ​​interest. Specifically, for the single target point cloud data obtained after the azimuth / elevation angle estimation, the false alarm suppression technology of machine learning (ML) such as SVM or RF algorithm can be combined to suppress non-ideal data such as false alarms generated by noise, stationary clutter or target multipath, that is, for the target point cloud data such as the pitch DBF&DoA output, continue to perform operations such as ML-based false-alarm suppression, clustering and / or ML-based target classification, where the clustering process can be agglomerative clustering or DBSCAN algorithms. At the same time, the point cloud data after clustering processing can also be input into a trained machine learning model (such as SVM or RF) to determine whether there is a person in the current processing. If there is, its location can be further determined, as well as the distinction and judgment of physical targets such as adults, children, and infants.

[0125] Figure 8In the embodiment shown, the influence of non-ideal factors such as noise, stationary clutter and target multipath can be effectively reduced, the difficulty of selecting regional parameters and the difficulty of complex regional judgment logic design can be effectively reduced, and point cloud information can be more effectively utilized, especially for applications in complex scenes, to obtain better target detection and distinction judgment performance.

[0126] Fig. 9 This is a flow chart of post-processing the target in combination with the DL algorithm in the embodiment of the present application. Fig. 9 As shown, it can be Figure 4 Based on the process shown, after extracting the data obtained from at least one of the operation steps such as multi-frame accumulation (store frame), frame Fourier transform (FrameFFT), non-coherent integration (Non-coherent integration), constant false alarm detection (CFAR), etc., a deep learning algorithm (DL, Deep Learning) is used to judge, locate and identify the target, for example, to identify whether the physical target is an adult or a child (Adult / Child Classification).

[0127] Specifically, the subsequent target distinction and judgment can be achieved by using 1D-FFT raw data as DL input (Access 1D-FFT data as raw input of DL), using 2D-FFT raw data between frames as DL input (Access 2D-FFT data as raw input of DL), using non-coherent integrated 2D FFT data as DL input (Access non-coherent integrated 2D FFT data as raw input of DL) and / or using SNR data as DL input (Access SNR data as raw input of DL). The raw data here can be complex data, modulus value of complex data, modulus square of complex data or other transformation forms, and can be values ​​in the linear domain, values ​​in the dB (log) domain or other calculation forms.

[0128] based on Fig. 9The process shown in the figure uses at least one of the above methods to extract multi-frame 1D-FFT data, multi-channel data after multi-frame FFT, and energy (power) or SNR data after multi-frame FFT and SISO-combine, and then inputs the extracted data into the constructed deep learning model (such as CNN or Transformer, etc.) to determine whether there is a target in the current processing, and to determine the current target type (such as determining whether the target is an adult or a child, etc.).

[0129] It should be noted that Fig. 9 The process steps and related descriptions shown in Figure 8 The process steps shown and the related descriptions may be compatible with each other; for example, Fig. 9 The Region HIST and subsequent modules in the process shown can be used Figure 8 In the process shown, the ML-based false alarm detection and its subsequent modules are partially or completely replaced with each other.

[0130] Fig.10 FIG. 1 is a flow chart of a target detection method combined with a DL algorithm in an embodiment of the present application. Fig.10 As shown, it can be Figure 4 On the basis of the process shown, after the inter-frame FFT (Frame FFT) processing, two-dimensional digital beamforming (2D-DBF) processing and DL-based target classification processing (DL-based Classification) operations are directly performed. That is, after the inter-frame FFT, CFAR processing is not performed, but 2D-DBF operations are performed on each region of interest (or unit interval) in the azimuth and elevation dimensions to obtain a number of RD spectra corresponding to the region of interest; subsequently, the above-mentioned RD spectra are input into the constructed deep learning model (DL) for target judgment. In some optional embodiments, the relationship between the above-mentioned region of interest and the RD spectra can also be one-to-many, that is, at least two RD spectra can be obtained based on one region of interest, and the specific number can be adjusted according to actual needs.

[0131] The above DL input can be linear or dB domain amplitude or power, and normalization processing and other operations can also be performed. At the same time, the deep learning model constructed above can be classified according to the input, and the categories distinguished can be: whether there is a target, the attributes of the target, the specific unit interval location (i.e., regional location) of each target, etc.

[0132] For example, in the application scenario where the rear area in the car is the area of ​​interest or the target area, and it is divided into 3 seat areas (interval units) and corresponding 3 aisle areas, after the inter-frame FFT, the center positions of the above 6 areas of interest (the areas of interest are also set as 3 seat areas) can be subjected to azimuth and elevation 2D-DBF to obtain 6 RD spectra (or 3 RD spectra) corresponding to the areas of interest, and at least part of the above 6 RD spectra (or 3 RD spectra) are input into the constructed deep learning model for target judgment and differentiation. For example, the constructed deep learning model performs classification based on the input to determine whether there is a living target in the above area of ​​interest. If there is a living target, it can be further determined whether the living target is an adult, a child, an infant or a pet, etc., and can further determine the specific location information of the living target, such as which seat or aisle area the living target is in.

[0133] Fig.10 In the embodiment shown, the target detection method is a processing flow based on dual drive of model data, which has low requirements for signal processing and can effectively avoid the steps of selecting signal processing parameters, regional parameters, designing CFAR logic and regional judgment logic, thereby greatly reducing the difficulty of solution implementation and design.

[0134] In some optional embodiments, for Figure 4-Figure 10 In the target detection method shown, all FFT processing can be windowed FFT, and can also be implemented by using SVA combined with FFT.

[0135] Fig.11 FIG. 1 is a flow chart of another target detection method combined with a DL algorithm in an embodiment of the present application. Fig.11 As shown, it can be Fig.10 Based on the process shown, after the range Fourier transform (Range FFT, also called 1D-FFT), 2D-DBF is performed first, and then frame data accumulation and inter-frame FFT processing are performed. That is, before the inter-frame FFT, the azimuth and elevation angles of the center position of the region of interest (such as the three-seat area) are firstly performed 2D-DBF. This embodiment can realize the solution of parallel processing of data processing and wave transmission time, thereby effectively reducing the processing time.

[0136] In some optional embodiments, for Figure 4-Figure 11 In the target detection method shown, the multi-frame sliding window FFT can also replace other time-frequency transform processing, such as short-time Fourier transform, fractional Fourier transform, etc., which can also achieve effective detection of the target of interest.

[0137] In some optional embodiments, for Figure 4-Figure 11In the target detection method shown, the multi-frame sliding window FFT can also be replaced by a high-pass FIR, a Comb FIR, or a filter optimized and synthesized according to an optimization function, so as to achieve the purpose of obtaining similar or even higher performance of multi-frame FFT processing with less hardware resources.

[0138] like Fig.10 , Fig.11 As shown, by combining 2D-FFT with DL recognition (DL-based Classification), the accuracy of target recognition can be effectively improved. At the same time, the 2D-FFT steps and the DL recognition steps can be flexibly set between various steps in signal processing according to actual needs, such as setting the 2D-FFT step after Range FFT, or after Frame FFT, etc.

[0139] Fig.12 FIG. 1 is a flow chart of another target detection method combined with a DL algorithm in an embodiment of the present application. Fig.12 As shown, it can be Fig.10 or Fig.11 On the basis of the process shown, that is, after the range-dimensional Fourier transform (Range FFT, also called 1D-FFT), the frame data is accumulated after 2D-DBF, and the DL-based target classification operation (ComplexDL-based Classification) is directly performed. That is, in the entire target detection process, no inter-frame FFT processing operation is performed, and 2D-DBF is directly performed on the center position of the area of ​​interest (such as the 3 seat area) after 1D-FFT to obtain the corresponding number of distance-frame maps (such as 3 range-frames), and then the above distance-frame maps are input into the constructed (for example, complex) deep learning model for target differentiation and judgment and other operations.

[0140] exist Fig.12 In the embodiment shown, since the FFT processing step between frames is avoided, the amount of calculation can be effectively reduced; at the same time, if a complex deep learning model is adopted, a higher processing gain can be obtained compared to FFT, thereby achieving the purpose of improving target detection performance; in addition, this embodiment also effectively shortens the processing level, which can facilitate its efficient scheduling operations, etc.

[0141] In some optional embodiments, for Figure 10-12 The DBF in the target detection method shown can be replaced by Capon, MUSIC, Esprint and other algorithms and their variants. At the same time, the beam synthesis at the center of the area of ​​interest and / or the antenna arrangement can be optimized and integrated to further improve the system target detection performance.

[0142] It should be noted that the embodiments of the present application can be mutually referenced and compatible without conflict, and the order and whether to set each functional module can be adjusted according to needs. At the same time, for a system with a BB module and an MCU module, when the system detects a target by emitting electromagnetic waves and receiving corresponding echo signals, each step in each target detection method of the embodiment of the present application can be configured to run in the BB module and / or the MCU module according to actual needs and the data processing capability and timeliness of the system operation. The relevant examples in the diagram can be used as a reference for some of the options.

[0143] In some optional embodiments, Figure 4-Figure 12 The target detection method related to the above is applicable to TDM-MIMO systems. For systems that do not use TDM-MIMO, different phase shifts can be preset when different Tx are transmitting. After the waveform of the directly transmitted waveform is digitally synthesized (phased array), the target detection method can be combined with the phase shift detection method to detect the target in the region of interest. Figure 10-12 The processing flow shown in the figure performs operations such as target detection and judgment.

[0144] In some other embodiments, for systems using a pulse system, a pulse compression system or an ultra-wideband system, directly using Figure 10-12 The method flow shown is used to detect targets, and the performance of target detection can be further improved by adopting the digital synthesis technology of the transmit beam. Among them, for the system of pulse compression system, 1D-FFT processing is not required, but the step of pulse compression needs to be added.

[0145] It should be noted that the above-mentioned processing methods corresponding to systems of different standards can also be applied to systems of mixed standards. For example, for a system with 6 transmitting antennas (i.e., 6 Tx), 3 of the transmitting antennas (3 Tx) can be TDM standards, and the other 3 transmitting antennas can be other standards. In the subsequent processing, the processing method corresponding to the above-mentioned standards can be used.

[0146] After the inter-frame FFT (Frame FFT) processing, two-dimensional digital beamforming (2D-DBF) processing and DL-based target classification processing (DL-based Classification) operations are directly performed. That is, after the inter-frame FFT, CFAR processing is not performed, but 2D-DBF operations are performed on the azimuth and elevation angles of each region of interest (or unit interval) to obtain the RD spectrum corresponding to the number of regions of interest; subsequently, the above-mentioned RD spectrum is input into the constructed deep learning model (DL) for target judgment. In some optional embodiments, if the inter-frame FFT method is not used, but FIR, short-time Fourier transform (STFT, Short-Time Fourier Transform) and other methods are used for processing, the above-mentioned RD spectrum corresponds to FIR as a one-dimensional spectrum output, and the corresponding STFT is a three-dimensional spectrum for time-frequency transform.

[0147] It should be noted that the electromagnetic waves in the embodiments of the present application may include radio waves and light waves. Radio waves include short waves, medium waves, long waves and microwaves, etc. Microwaves include centimeter waves (i.e. electromagnetic waves in the range of 3GHz to 30GHz, such as electromagnetic waves in the range of 3.1GHz to 10.6GHz, electromagnetic waves in the frequency band of 24GHz, etc.) and millimeter waves (i.e. electromagnetic waves in the range of 30GHz to 300GHz, such as electromagnetic waves in the frequency band of 60GHz, electromagnetic waves in the frequency band of 77GHz (such as 77GHz to 81GHz, etc.)), and light waves may include ultraviolet rays, visible light, infrared rays, and lasers, etc., among which the electromagnetic wave frequency band of lasers is (3.846 to 7.895)*10^5GHz, i.e., lasers are included in part of the frequency bands of ultraviolet rays and visible light.

[0148] An embodiment of the present application also provides another target detection method, including: performing distance-dimensional FFT processing based on the echo signal to obtain 1D-FFT data; and performing at least one of multi-frame joint processing, deep learning target detection, deep learning target positioning, and deep learning target recognition on the 1D-FFT data to achieve target detection in the area of ​​interest.

[0149] In an exemplary embodiment, when the digital data signal is jointly processed in multiple frames, the method includes: accumulating frame data of the 1D-FFT data until a preset amount of data is accumulated, performing FFT processing between frames to obtain an RD spectrum; and detecting targets in a region of interest based on the RD spectrum.

[0150] In an exemplary embodiment, for a multi-channel application scenario, the detection of targets in the region of interest based on the RD spectrum includes: after incoherent accumulation of the RD spectrum of each channel, continuous constant false alarm processing based on noise estimation, direction of arrival estimation, and target judgment, positioning and recognition operations based on deep learning.

[0151] In an exemplary embodiment, the constant false alarm processing is to perform DAE CFAR processing on the azimuth angle and the elevation angle respectively.

[0152] In an exemplary embodiment, the deep learning target recognition of the 1D-FFT data includes: based on extracting at least part of the multi-frame accumulated data, the inter-frame Fourier transform data, the incoherently accumulated data and / or the constant false alarm detection data, using a deep learning algorithm to perform a target recognition operation.

[0153] In an exemplary embodiment, for a multi-channel application scenario, the detection of targets in the region of interest based on the RD spectrum includes: performing constant false alarm processing on the RD spectrum of each channel, continuing with binary integration processing and direction of arrival estimation in the channel domain, and then performing target judgment, positioning and recognition operations based on deep learning.

[0154] In an exemplary embodiment, the detection of the target in the region of interest based on the RD spectrum includes: performing two-dimensional digital beam synthesis on the RD spectrum, and continuing the DL-based target classification processing to achieve target judgment, positioning and recognition operations.

[0155] In an exemplary instance, deep learning target detection, deep learning target positioning and / or deep learning target recognition are performed on the 1D-FFT data, including: obtaining target point cloud data based on the 1D-FFT data; and performing ML-based false alarm suppression, clustering processing and / or ML-based target classification processing on the target point cloud data to achieve judgment, positioning and / or recognition operations on the target in the area of ​​interest.

[0156] In an exemplary instance, the method further includes: after performing two-dimensional digital beam synthesis on the 1D-FFT data, continuing to accumulate frame data until a preset amount of data is accumulated, performing inter-frame FFT processing to obtain an RD spectrum; and performing DL-based target classification processing based on the RD spectrum to achieve target judgment, positioning and recognition operations.

[0157] An embodiment of the present application further provides a method for target detection, including: performing distance-dimensional FFT processing based on the echo signal to obtain 1D-FFT data; performing two-dimensional digital beam synthesis on the 1D-FFT data, and continuing to accumulate frame data; and when a preset amount of data has been accumulated, performing DL-based target classification processing to achieve target judgment, positioning and identification operations.

[0158] The embodiment of the present application also provides an electromagnetic wave sensor, including: a signal transmitting module, configured to transmit electromagnetic waves for target detection; a signal receiving module, configured to receive echoes formed by reflection and / or scattering of the electromagnetic waves; and a processing module, configured to perform signal and data processing on the echoes to achieve operations such as target judgment, positioning and / or identification; wherein the processing module is also configured to implement any target detection method provided in the embodiment of the present application. In one embodiment, the processing module includes a baseband unit and an MCU unit, the baseband unit can be configured to implement operations before clustering analysis of the original point cloud in the above-mentioned target detection method, and the MCU unit can be configured to implement the above-mentioned target detection method.

[0159] The embodiment of the present application also provides an integrated circuit, which may include a radio frequency module, an analog signal processing module and a digital signal processing module connected in sequence; the radio frequency module is used to generate a radio frequency transmission signal and receive an echo signal; the analog signal processing module is used to down-convert the echo signal to obtain an intermediate frequency signal; the digital processing module is used to perform analog-to-digital conversion on the intermediate frequency signal to obtain a digital signal; and based on the target detection method in the embodiment of the present application, the digital signal is processed to achieve the purpose of target detection, for example, the integrated circuit may be a millimeter wave radar chip (chip or die). Among them, the above-mentioned digital processing module may include sub-units such as a BB unit and an MCU unit, and each sub-unit may be configured to execute each corresponding step in the target detection method in the above-mentioned embodiment.

[0160] In some optional embodiments, the integrated circuit may be an antenna-in-package (AiP) chip structure, an antenna-on-package (AoP) chip structure, an antenna-on-chip (AoC) chip structure, or a Radiator through Package (RoP) chip structure. Optionally, the RoP chip structure may be a radiator structure (Radiator) provided on the chip package (Package), and an air waveguide structure may be formed around the radiator structure with solder balls, that is, the radio frequency (RF) signal generated by the chip may be transmitted to an external antenna through the above-mentioned radiator structure, the air cavity waveguide structure formed by the solder balls, and the air waveguide built into the PCB board, so as to radiate to the target area.

[0161] According to other embodiments of the present application, an electromagnetic wave sensor is also proposed. The electromagnetic wave sensor may include an antenna, and an integrated circuit as described above. The integrated circuit is electrically connected to the antenna for receiving and transmitting electromagnetic wave signals. For example, the electromagnetic wave sensor may include: a carrier, an integrated circuit and an antenna as described in any of the above embodiments, and the integrated circuit may be arranged on the carrier; the antenna may be arranged on the carrier, or the integrated circuit may be integrated with the integrated circuit as an integrated device arranged on the carrier (that is, the antenna may be an antenna provided in an AiP, AoP or AoC structure at this time); the integrated circuit is connected to the antenna (that is, the sensor chip or integrated circuit is not integrated with an antenna at this time, such as a conventional SoC, etc.), for receiving and transmitting electromagnetic wave signals. The carrier may be a printed circuit board PCB, and the corresponding transmission line may be a PCB trace.

[0162] An embodiment of the present application provides a device, which may include: a device body; and an electromagnetic wave sensor as described above, which is arranged on the device body; wherein the electromagnetic wave sensor is used for target detection and / or communication to provide reference information for the operation of the device body.

[0163] The embodiment of the present application also provides a terminal device, which can be expressed in the form of a general-purpose computing device. The components of the terminal device may include, but are not limited to: at least one processing unit, at least one storage unit, a bus connecting different system components (including a storage unit and a processing unit), a display unit, etc. Among them, the storage unit stores a program code, and the program code can be executed by the processing unit so that the processing unit executes the method described in this specification according to various exemplary embodiments of the present application. The storage unit may include a readable medium in the form of a volatile storage unit, such as a random access memory unit (RAM) and / or a cache memory unit, and may further include a read-only memory unit (ROM).

[0164] The storage unit may also include a program / utility having a set (at least one) of program modules, such program modules including but not limited to: an operating system, one or more application programs, other program modules, and program data, each of which or some combination may include an implementation of a network environment.

[0165] The bus may represent one or more of several types of bus structures, including a memory unit bus or memory unit controller, a peripheral bus, an accelerated graphics port, a processing unit, or a local bus using any of a variety of bus architectures.

[0166] The terminal device can also communicate with one or more external devices (such as keyboards, pointing devices, Bluetooth devices, etc.), and can also communicate with one or more devices that enable users to interact with the terminal device, and / or communicate with any device (such as routers, modems, etc.) that enables the terminal device to communicate with one or more other computing devices. This communication can be carried out through an input / output (I / O) interface. In addition, the terminal device can also communicate with one or more networks (such as local area networks (LANs), wide area networks (WANs) and / or public networks, such as the Internet) through a network adapter. The network adapter can communicate with other modules of the terminal device through a bus. It should be understood that, although not shown in the figure, other hardware and / or software modules can be used in conjunction with the terminal device, including but not limited to: microcode, device drivers, redundant processing units, external disk drive arrays, RAID systems, tape drives, and data backup storage systems, etc.

[0167] For example, the terminal device in the embodiment of the present application may also include: a device body; and an electromagnetic wave sensor as described in any of the above embodiments arranged on the device body; wherein the electromagnetic wave sensor can be used to realize functions such as target detection and / or wireless communication.

[0168] Specifically, based on the above embodiments, in an optional embodiment of the present application, the electromagnetic wave sensor can be arranged outside the device body, or arranged inside the device body, and in other optional embodiments of the present application, the electromagnetic wave sensor can also be arranged partially inside the device body and partially outside the device body. The embodiments of the present application are not limited to this, and the specific situation can be determined.

[0169] In an optional embodiment, the device body can be a component or product used in fields such as smart cities, smart homes, transportation, smart homes, consumer electronics, security monitoring, industrial automation, in-cabin detection (such as smart cockpits), medical devices, and health care. For example, the device body can be intelligent transportation equipment (such as cars, bicycles, motorcycles, ships, subways, trains, etc.), security equipment (such as cameras), liquid level / flow rate detection equipment, smart wearable devices (such as bracelets, glasses, etc.), smart home devices (such as sweeping robots, door locks, televisions, air conditioners, smart lights, etc.), various communication devices (such as mobile phones, tablet computers, etc.), and gates, smart traffic lights, smart signs, traffic cameras, and various industrialized robotic arms (or robots), etc. It can also be various instruments for detecting life characteristic parameters and various devices equipped with the instruments, such as life characteristic detection in car cabins, indoor personnel monitoring, smart medical equipment, consumer terminal equipment, etc.

[0170] The embodiment of the present application also provides a non-transitory computer-readable storage medium on which computer-readable instructions are stored. When the instructions are executed by a processor, the processor executes the target detection method as described above.

[0171] Through the description of the above implementation methods, it is easy for those skilled in the art to understand that the example implementation methods described here can be implemented by software or by combining software with necessary hardware. The technical solution according to the implementation method of the present application can be embodied in the form of a software product, which can be stored in a non-volatile storage medium (which can be a CD-ROM, a USB flash drive, a mobile hard disk, etc.) or on a network, including several instructions to enable a computing device (which can be a personal computer, a server, or a network device, etc.) to execute the above method according to the implementation method of the present application.

[0172] The software product may use any combination of one or more readable media. The readable medium may be a readable signal medium or a readable storage medium. The readable storage medium may be, for example, but not limited to, a system, device or device of electricity, magnetism, light, electromagnetic, infrared, or semiconductor, or any combination thereof. More specific examples (non-exhaustive list) of readable storage media include: an electrical connection with one or more wires, a portable disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination thereof.

[0173] Computer readable storage media may include data signals propagated in baseband or as part of a carrier wave, wherein readable program codes are carried. Such propagated data signals may take various forms, including but not limited to electromagnetic signals, optical signals, or any suitable combination thereof. The readable storage medium may also be any readable medium other than a readable storage medium, which may send, propagate, or transmit a program for use by or in conjunction with an instruction execution system, device, or device. The program codes contained on the readable storage medium may be transmitted using any suitable medium, including but not limited to wireless, wired, optical cable, RF, etc., or any suitable combination thereof.

[0174] Program code for performing the operations of the present application may be written in any combination of one or more programming languages, including object-oriented programming languages ​​such as Java, C++, etc., and conventional procedural programming languages ​​such as "C" or similar programming languages. The program code may be executed entirely on the user computing device, partially on the user device, as a separate software package, partially on the user computing device and partially on a remote computing device, or entirely on a remote computing device or server. In the case of a remote computing device, the remote computing device may be connected to the user computing device through any type of network, including a local area network (LAN) or a wide area network (WAN), or may be connected to an external computing device (e.g., via the Internet using an Internet service provider).

[0175] The computer-readable medium carries one or more programs. When the one or more programs are executed by a device, the computer-readable medium implements the aforementioned functions.

[0176] Those skilled in the art will appreciate that the above modules can be distributed in the device according to the description of the embodiment, or can be changed accordingly and only used in one or more devices different from the embodiment. The modules of the above embodiments can be combined into one module, or further divided into multiple sub-modules.

[0177] According to an embodiment of the present application, a computer program is proposed, including a computer program or an instruction, which, when executed by a processor, can execute the method described above. In an optional embodiment, the above-mentioned integrated circuit can be a millimeter wave radar chip. The type of digital function module in the integrated circuit can be determined according to actual needs. For example, in a millimeter wave radar chip, the data processing module can be used for functions such as distance-dimensional Doppler transform, velocity-dimensional Doppler transform, constant false alarm detection, wave direction detection, point cloud processing, etc., to obtain information such as the distance, angle, speed, height, micro-Doppler motion characteristics, shape, size, surface roughness and dielectric properties of the target.

[0178] It should be noted that radio devices can achieve functions such as target detection and / or communication by transmitting and receiving radio signals to provide the device body with detection target information and / or communication information, thereby assisting or even controlling the operation of the device body.

[0179] For example, when the above-mentioned device body is applied to an advanced driver assistance system (i.e., ADAS), the radio device (such as millimeter-wave radar) as a vehicle-mounted sensor can assist the ADAS system to realize application scenarios such as adaptive cruise control, automatic braking assistance (i.e., AEB), blind spot detection warning (i.e., BSD), assisted lane change warning (i.e., LCA), reversing assistance warning (i.e., RCTA), parking assistance, rear vehicle warning, collision avoidance, pedestrian detection, and in-cabin liveness detection (i.e., CPD).

[0180] Fig.13 This is a flow chart of a radar data processing embodiment in a target detection method in an embodiment of the present application, combined with Fig.19 ,include:

[0181] Step 1300: Filter out the first DC component from the ADC data.

[0182] In one exemplary embodiment, step 1300 may include:

[0183] For example, in baseband processing (BB), based on the obtained ADC data, the average value of the sampled data of each receiving channel (RX channel) is calculated along the fast time dimension and the calculated average value is used as the first DC component of the channel; for all the sampled data of each RX channel, the first DC component of the channel is subtracted respectively to obtain the data after DC filtering.

[0184] Analog-to-digital converter (ADC) data refers to the digital signal obtained by analog-to-digital conversion of the analog signal received from the radar receiving antenna. ADC data is the digital signal collected by the radar receiving channel to each RX channel during each linear frequency modulation signal (chirp), that is, the data collected from each RX channel for each chirp.

[0185] In some embodiments, assuming that N sampling points are collected in the fast time dimension in one chirp, then the average value of the sampled data of each RX channel along the fast time dimension is calculated as the first DC component of the channel. In this embodiment, assuming that x ij Indicates the signal strength of the i-th RX channel at the j-th sampling point. The first DC component DC i The calculation of is shown in formula (1):

[0186]

[0187] Then, the data of the i-th RX channel after filtering out the first DC component is x′ ij =x ij -DC i .

[0188] For example, suppose there are two RX channels, each collecting data from 4 sampling points. The data is as follows: RX1 channel data: [10, 12, 11, 13]; RX2 channel data: [20, 22, 21, 23];

[0189] According to formula (1), the first DC component of RX1 channel can be obtained: The first DC component of the RX2 channel:

[0190] The data of RX1 channel after DC component is filtered out is: [10-11.5, 12-11.5, 11-11.5, 13-11.5] = [-1.5, 0.5, -0.5, 1.5]; the data of RX2 channel after DC component is filtered out is: [20-21.5, 22-21.5, 21-21.5, 23-21.5] = [-1.5, 0.5, -0.5, 1.5].

[0191] During the reception and processing of radar signals, DC components may be introduced into the received signals due to hardware imbalance or other system deviations. In the embodiment of the present application, the DC component of the data of each RX channel is filtered out by filtering the DC component of the downlink ADC data before the distance and speed processing in the baseband processing stage, so that the subsequent signal processing is more accurate and effective.

[0192] Step 1301: After the data of filtering out the first DC component is windowed, a one-dimensional fast Fourier transform is performed to obtain distance dimension data and store the data.

[0193] In an exemplary embodiment, in order to reduce spectrum leakage, the data after the first DC filter of each RX channel is windowed in the embodiment of the present application. For example, the data after the DC filter of each RX channel can be windowed using any of the following window functions: Hanning window, Blackman window, etc. In one embodiment, assuming that the window function is w j ,but

[0194] ″′

[0195] The windowed data is: x ij =x ij ·w j , so that the signal gradually decays to zero at the boundary of the time domain, thereby reducing the truncation effect and spectrum leakage.

[0196] In an exemplary embodiment, a one-dimensional FFT (1D-FFT) is performed on the windowed data to convert the time domain signal into a frequency domain signal to obtain distance dimension data. The distance dimension data is data extracted from the received signal reflecting the distance characteristics of the target object.

[0197] For each RX channel, the data after 1D-FFT is: Among them, k represents the frequency index, corresponding to the distance unit.

[0198] In an exemplary embodiment, the distance dimension data obtained by fast Fourier transformation in step 1301 can be moved from the BB to the static random access memory (SRAM) cache of the microcontroller unit (MCU) through direct memory access (DMA). DMA is a technology for transferring data between memory and peripherals without passing through the CPU. Using DMA greatly improves data transmission efficiency and reduces CPU burden.

[0199] In the embodiment of the present application, by moving the obtained distance dimension data to the SRAM cache of the CPU, fast access and processing are achieved, while avoiding the CPU being busy with data transmission, thereby improving the overall efficiency of the system.

[0200] Step 1302: Return to step 100 to continue filtering out the first DC component of the ADC data until the accumulated distance dimension data reaches a preset number of frames.

[0201] In an exemplary embodiment, multiple frames of distance dimension data of a preset number of frames may be moved from the CPU SRAM to the BB via DMA to process the multiple frames of data.

[0202] In an exemplary embodiment, the preset number of frames may be, for example, 256 frames, 128 frames, 56 frames, or 32 frames.

[0203] Step 1303: Filter out the second DC component from the distance dimension data of a preset number of frames.

[0204] In an exemplary embodiment, step 1303 may include:

[0205] For the preset frame number N frames The complex average value of the data along the frame dimension (i.e., the time dimension) is calculated to obtain the average values ​​of different RX channels and different distance units as the second DC component of the corresponding RX channel For the distance dimension data of each RX channel, the second DC component of the channel is subtracted to obtain the data after DC filtering. Such processing increases the system's sensitivity to changes in weak signals. For example, in in-car baby detection, it is sometimes necessary to detect very weak physiological signals or dynamic changes such as a baby's breathing and heartbeat. By removing the second DC component in the embodiment of the present application, these signals are made clearer and detectable, thereby improving the reliability and accuracy of detection.

[0206] Among them, the second DC component As shown in formula (2):

[0207]

[0208] Then, the data of the i-th RX channel after filtering out the second DC component is

[0209] Step 1304: After filtering out the second DC component, window processing is performed on the data, and then two-dimensional fast Fourier transform is performed to obtain range Doppler data.

[0210] In an exemplary embodiment, in order to reduce spectrum leakage, the data after the second DC component is filtered out is windowed along the frame dimension in the embodiment of the present application. For example, the DC filtered out data of each RX channel can be windowed using any of the following window functions: Hanning window, Blackman window, etc. In one embodiment, assuming that the window function is w k , then the windowed data is:

[0211] In an exemplary embodiment, a two-dimensional FFT (2D-FFT) is performed on the windowed data. For each RX channel, the data after 2D-FFT is: Among them, k represents the frequency index, corresponding to the distance unit; m represents the frame index, indicating the position in the time dimension; r represents the distance unit index, indicating the position in the distance dimension; v is the Doppler frequency index (velocity unit), indicating the position in the Doppler dimension (velocity dimension). Through 2D-FFT transformation, S(r,v) can be obtained, which represents the frequency domain signal corresponding to each distance unit and Doppler frequency unit. Subsequently, these signals can be used to extract the distance and velocity information of the target.

[0212] The embodiment of the present application is different from the slow time processing between chirps of radar signal processing in the related art. Instead, it accumulates multiple frames of data and performs inter-frame FFT processing in the form of a sliding window. On the one hand, in the embodiment of the present application, the superposition of multiple frames of data can offset the randomness of noise, and the accumulation of multiple frames of data can increase the signal-to-noise ratio (SNR) of the processing, making the target signal more significant, thereby improving the strength and clarity of the signal, and effectively improving the radar system's ability to detect weak signal targets; on the other hand, the use of sliding windows for FFT processing on the accumulated multiple frames of data allows local spectrum analysis of time series signals, which can provide more accurate spectrum information in dynamic scenes and more accurately estimate the speed information of the target.

[0213] Step 1305: Perform incoherent accumulation based on the acquired range-dimensional Doppler data, estimate the noise floor of each range unit of the incoherently accumulated data, and then perform CFAR processing.

[0214] In an exemplary embodiment, performing incoherent accumulation based on the acquired range-dimensional Doppler data may include:

[0215] The obtained range-dimensional Doppler data are subjected to multi-channel processing to obtain a multi-channel range-Doppler map; and the multi-channel range-Doppler map is subjected to incoherent accumulation.

[0216] In one embodiment, performing non-coherent accumulation of multi-channel range-Doppler maps (RD maps) may include calculating the power P(r, v) or amplitude A(r, v) as shown in formula (3) and formula (4):

[0217]

[0218] Among them, N t Indicates the number of transmission channels; N r represents the number of receiving channels; P(r,v) represents the power at the rth distance unit and the vth Doppler unit; A(r,v) represents the amplitude at the rth distance unit and the vth Doppler unit; s(r,t,a,v) represents the complex value of the signal echo on the tth transmitting channel of the rth distance unit and the vth Doppler unit and the ath receiving channel after the inter-frame FFT processing in step 1304.

[0219] In an exemplary embodiment, the noise floor of each distance unit of the incoherently accumulated data can be estimated by the formula n′ i =min(n i ,αn g ) The noise floor of the incoherently accumulated data is estimated and corrected to accurately estimate the noise level and ensure the accuracy of CFAR detection. i represents the corrected noise floor estimate of the ith distance unit, n irepresents the original noise floor estimate of the i-th range unit; n g represents the noise floor estimate of the last range cell of interest; α is the adjustment factor.

[0220] In an exemplary embodiment, the constant false alarm rate (CFAR) processing may be a non-coherent constant false alarm rate (NR-CFAR) detection method. In an embodiment, the CFAR processing may include using a CFAR algorithm to perform possible target point detection, identifying possible target signal points and performing peak detection to obtain CFAR detection points, thereby achieving detection of target points in a noise background.

[0221] Step 1306: Determine the position and motion state of the possible target in space according to the CFAR detection points obtained after CFAR processing to obtain an original point cloud.

[0222] In one illustrative example, step 1306 may include:

[0223] Performing azimuth digital beamforming (DBF) and angle of arrival (DoA) estimation on the CFAR detection points to obtain the azimuth angle of the possible target, and performing elevation DBF and DoA estimation on the CFAR detection points to obtain the elevation angle of the possible target;

[0224] According to the azimuth and pitch angle of the possible target, as well as the target distance and speed information detected by CFAR, the position information and motion state (speed v) of the possible target in space are calculated;

[0225] The distance, azimuth, pitch angle and speed information of possible targets in the polar coordinate system are converted into the four-dimensional information (x, y, z, v) of the target in the Cartesian coordinate system, where x can represent the horizontal direction of the east-west direction, y can represent the horizontal direction of the north-south direction, and z can represent the vertical direction of the up-down direction.

[0226] In one embodiment, the original point cloud in the present application is a point cloud obtained after inter-frame FFT processing.

[0227] Step 1307: The obtained position and motion state of the possible target in space are transmitted to the controller to perform cluster analysis on the original point cloud, and the target is tracked and the target point cloud is generated according to the clustering result.

[0228] In an exemplary embodiment, the data of the CFAR detection point obtained in step 106 may be transferred from the baseband processor to the SRAM of the MCU via DMA.

[0229] In an exemplary embodiment, the original point cloud may be clustered using a density-based spatial clustering of applications with noise (DBSCAN) algorithm using x, y, z information in the four-dimensional information (x, y, z, v) of the possible target to group the CRAF detection points (i.e., possible targets) into target point clusters. In an embodiment, clustering analysis of the original point cloud may include:

[0230] Select points that meet the requirements of core objects from possible targets, that is, core objects that contain at least MinPts points in the ε neighborhood;

[0231] The core points in each other's ε neighborhood are connected to form a point cluster, and the points in the ε neighborhood of the core point of the cluster are also added to the point cluster, eventually forming multiple point clusters, and the center point of each cluster point cloud is used as the cluster center point.

[0232] In an exemplary embodiment, the clustering algorithm may also be a partitioning clustering algorithm k-means, a density-based algorithm OPTICS, a hierarchical clustering algorithm Agglomerative, and the like.

[0233] In some embodiments, the clustered raw data may also be various combinations of the four-dimensional information (x, y, z, v) of the possible target, the x, y information in the four-dimensional information (x, y, z, v) of the possible target, the x, z information in the four-dimensional information (x, y, z, v) of the possible target, the y, z information in the four-dimensional information (x, y, z, v) of the possible target, and the x, y, v information in the four-dimensional information (x, y, z, v) of the possible target.

[0234] In some embodiments, the center point of each cluster point cloud can be the average value (center of mass) of each cluster point cloud, or the median value of each cluster point cloud can be taken as the cluster center point, which can better reduce the impact of large deviations in each cluster point cloud.

[0235] In an exemplary embodiment, performing target tracking and generating a target point cloud according to the clustering result in step 1307 may include:

[0236] The trajectory of the previous frame in each track is used as the predicted value of the current frame;

[0237] Calculate the distance between the predicted value of each track and all cluster center points of the current frame;

[0238] The cluster center points whose calculated distance D is less than the preset threshold are considered to match the track and are added to the corresponding track, which means that these cluster center points are considered to be associated with the track and may be the new position of the track in the current frame or the continued track; for the unmatched cluster center points, they will be added to the candidate tracks for track start determination, that is, these candidate tracks may be newly appeared targets or the starting part of the original track, and further observation and analysis are needed to determine whether they form a valid track;

[0239] Update the tracks that have matched the cluster center points; in one embodiment, this may include: for existing tracks, if no matching cluster center points are found in the current frame, then the end flag values ​​of these tracks will be increased. This end flag can be used to determine whether the track has ended or is temporarily interrupted for subsequent track management and analysis. For tracks that have not matched the cluster center points, check whether the value of their end flags exceeds the preset threshold, and if so, remove these tracks; add candidate tracks that meet the MN judgment to the formal tracks, and the MN judgment criteria are: if the candidate track has N frames that can be associated with the cluster center point within the past M frames, it means that the judgment is successful;

[0240] The newly added cluster center points in all tracks are extracted, and the original point cloud and track number of the single cluster corresponding to each cluster center point are output to generate the target point cloud.

[0241] In step 1307 of the embodiment of the present application, the point cloud data is divided into different clusters by clustering, and the centroid (center point) of each cluster is calculated, and then the target track is updated and managed based on the association between these centroids and the trajectory of the previous frame to achieve trajectory prediction and matching.

[0242] In an exemplary embodiment, the distance calculation method may include, but is not limited to, calculation based on the NN method, or Chebyshev distance, or Manhattan distance (L1 distance), etc.

[0243] In an exemplary embodiment, the Euclidean distance between the track prediction value and each cluster center point is calculated based on the NN algorithm to represent the correlation between the center point and the track. In this embodiment, only the x, y, z information in the four-dimensional information (x, y, z, v) is used. In some embodiments, the v information may also be added, that is, the four-dimensional information is used for association. In other embodiments, for radars without pitch angle measurement capabilities, only the x, y, v information in the four-dimensional information (x, y, z, v) or the x, y information in the four-dimensional information (x, y, z, v) may be used for association.

[0244] The radar data processing method provided in the embodiment of the present application uses a multi-frame joint processing technology, adopts a new inter-frame accumulation method, and combines the detected point cloud clustering and track processing methods to effectively filter out false alarm points and multipath target points, thereby improving the quality of the detected point cloud, thereby providing a guarantee for improving the detection, measurement or recognition performance, and further effectively improving the detection rate, reducing the number of false alarm targets, and thus improving the accuracy of target detection.

[0245] Furthermore, the radar data processing method provided in the embodiment of the present application, while effectively improving the detection rate and reducing the number of false alarm targets, also greatly improves the accuracy of angle estimation and other effects, thereby enabling special targets or weak targets such as infants and young children in the cabin to be accurately detected, thereby realizing applications such as CPD and SBR.

[0246] Step 1308: Detect the target in the detection area using the target point cloud.

[0247] In one exemplary embodiment, in combination Fig. 20 , Fig. 20 In the example of performing two-dimensional digital beamforming (2D-DBF) processing directly after frame FFT processing and then performing clustering and tracking processing, step 1308 may include:

[0248] Divide the entire detection area into several preset areas;

[0249] Count the target points detected in each frame according to the preset area;

[0250] According to the number of times the target appears in the preset area, it is determined whether there is a target in the detection area and the specific area where it is located.

[0251] The target detection method provided by the embodiment of the present application realizes target detection and identification in spatial areas such as car cabins, indoors, and factories, and has a better detection effect on the detection of infants (targets) in the car. The target detection method provided by the embodiment of the present application is particularly suitable for the scene of human target detection and positioning in the cabin with micro-motion and low speed.

[0252] FIG14(a) is a schematic diagram of the point cloud result in the related art under the first test scenario, and FIG14(b) is a schematic diagram of the point cloud result after processing by the embodiment of the present application under the first test scenario. The first test scenario is a baby in a car. At a certain moment, the car is shaken. As shown in FIG14(a), there are some interferences in the original radar data, such as occasional interference points at about 2.6m in the y direction. At 32 to 35 frames, the target position is rapidly offset due to the shaking of the car. At this time, the point cloud shown in FIG14(a) has a maximum offset of about 0.5m in the x and z directions. These interferences may cause errors in the determination of the target position, and as shown in FIG14(b), these interference points are filtered out after the tracking data is processed, which effectively improves the quality of the point cloud.

[0253] Figure 15(a) is a schematic diagram of the point cloud result in the related art under the second test scenario, and Figure 15(b) is a schematic diagram of the point cloud result after processing by the embodiment of the present application under the second test scenario. The second test scenario is that the baby is in the car and the door is opened and closed. In this scenario, the door opening and closing operations will cause the car body to shake and the radar will receive the reflected signal of the target outside the car. In the y direction, as shown in Figure 15(a), the point cloud will have some intermittent interference points more than 1.5 meters away, which can be effectively filtered out after the tracker data processing as shown in Figure 15(b).

[0254] FIG16(a) is a schematic diagram of the point cloud result in the related art under the third test scenario, and FIG16(b) is a schematic diagram of the point cloud result after processing by the embodiment of the present application under the third test scenario. The third test scenario is the detection of adult seat A. As shown in FIG16(a), the x, y, z graph shows the spatial coordinates (distance / meter) of the point in the point cloud in a series of frames, and the v graph shows the speed (Hz) of the point in the frame. There are several clusters of data points, accompanied by some noise and outliers, and the distribution of data points is relatively scattered. As shown in FIG16(b), the x, y, z graph shows the spatial coordinates (distance / meter) of the point in the point cloud in a series of frames, but some filtering or tracking algorithms have been applied, and the v graph shows the speed (Hz) of the point in the frame after filtering. Compared with the original point cloud data, the data points appear to be more concentrated and structured. The filtering process reduces noise and outliers, making the clustering of data points clearer, and the structure and clustering in the filtered point cloud data are more obvious, which indicates that the accuracy in representing the position and speed of the points has been improved.

[0255] Figure 17(a) is a schematic diagram of the point cloud result in the related art under the fourth test scenario, and Figure 17(b) is a schematic diagram of the point cloud result after processing by the embodiment of the present application under the fourth test scenario, and the fourth test scenario is that the baby is not covered with a blanket. In this most common scenario where the car is stationary and only contains the baby falling in the car, the point cloud shown in Figures 16(a) and 17(a) will produce some false alarm points, especially sporadic false alarms in the y direction, while the tracking data processing shown in Figures 16(b) and 17(b) can effectively filter out the false alarm points.

[0256] The target detection method provided by the embodiment of the present application effectively filters out false alarm points and multipath target points, thereby realizing a solution for target detection and identification in spatial areas such as automobile cabins, indoors, and factory buildings. While effectively improving the detection rate and reducing the number of false alarm targets, the accuracy of angle estimation and other effects are greatly improved, thereby realizing accurate detection of special targets or weak targets such as infants and young children in the cabin, and realizing applications such as CPD and SBR. Specifically, performance evaluation is carried out through a large number of actual measurement experiments, and the results show that the target detection method provided by the embodiment of the present application effectively filters out false alarm points and multipath points, effectively improves the detection rate, reduces the number of false alarm targets, and thus improves the accuracy of target detection.

[0257] Fig.18 FIG. 1 is a schematic diagram of the structure of a radar data processing device in a target detection method in an embodiment of the present application. Fig.18 As shown, it may include: a first DC component processing module, a first processing module, a cache module, a second DC component processing module, a second processing module, a third processing module, a first acquisition module, a second acquisition module and a detection module; wherein,

[0258] A first DC component processing module, used for filtering out a first DC component from ADC data;

[0259] A first processing module, configured to perform windowing processing on the data after filtering out the first DC component, obtain distance dimension data through a one-dimensional fast Fourier transform, and store the data in a cache module;

[0260] A second DC component processing module is used to read the distance dimension data of the preset number of frames stored in the cache module when the distance dimension data stored in the cache module reaches a preset number of frames, and filter out the second DC component from the read distance dimension data of the preset number of frames;

[0261] A second processing module is used for performing windowing processing on the data after filtering out the second DC component, and then obtaining range Doppler data through a two-dimensional fast Fourier transform;

[0262] The third processing module is used to perform incoherent accumulation according to the obtained range-dimensional Doppler data, and to estimate the noise floor of each range unit of the incoherently accumulated data and then perform CFAR processing;

[0263] The first acquisition module is used to determine the position and motion state of the possible target in space according to the CFAR detection points obtained after CFAR processing to obtain the original point cloud;

[0264] The second acquisition module is used to transmit the acquired position and motion state of the possible target in space to the controller to perform cluster analysis on the original point cloud, track the target according to the clustering result, and generate the target point cloud.

[0265] In an exemplary embodiment, Fig. 9 As shown, the first DC component processing module, the first processing module, the second DC component processing module, the second processing module, the third processing module, and the first acquisition module are arranged in the BB unit of the radar device, and the cache module, the second acquisition module, and the detection module are arranged in the MCU of the radar device.

[0266] In an exemplary embodiment, the first DC component processing module may be a Range DC Removal module, which may be used to:

[0267] In BB processing, based on the obtained ADC data, the average value of the sampled data of each RX channel is calculated along the fast time dimension and the calculated average value is used as the first DC component of the channel; for all the sampled data of each RX channel, the first DC component of the channel is subtracted respectively to obtain the data after DC filtering.

[0268] In an exemplary embodiment, the first processing module may be a range FFT module, which may be used to:

[0269] The data after the first DC filtering of each RX channel is windowed; 1D-FFT is performed on the windowed data to convert the time domain signal into a frequency domain signal to obtain distance dimension data; and the obtained distance dimension data is moved from the BB to the cache module of the MCU through DMA.

[0270] In one embodiment, the cache unit may include, but is not limited to, SRAM.

[0271] In an exemplary embodiment, the second DC component processing module may be a Doppler DC Removal module, which may be used to:

[0272] Read the distance dimension data of the preset number of frames stored in the cache module; frames The complex average value of the data along the frame dimension (i.e., the time dimension) is calculated to obtain the average values ​​of different RX channels and different distance units as the second DC component of the corresponding RX channel For the distance dimension data of each RX channel, the second DC component of the channel is subtracted to obtain the data after DC filtering.

[0273] In an exemplary embodiment, the second processing module may be a frame FFT module, which may be used to:

[0274] The data after the second DC component is filtered out is windowed along the frame dimension; 2D-FFT is performed on the windowed data of each RX channel to obtain a frequency domain signal corresponding to each range unit and Doppler frequency unit.

[0275] In an exemplary embodiment, the third processing module may include: a non-coherent integration module, a noise variance estimation (NVE) module, and a CFAR detection with peak selection module; wherein,

[0276] The incoherent accumulation module is used to perform multi-channel processing on the acquired range-dimensional Doppler data to obtain a multi-channel range-Doppler map; and perform incoherent accumulation on the multi-channel range-Doppler map;

[0277] Noise variance estimation module, used to estimate and correct the noise floor of the data after incoherent accumulation;

[0278] The CFAR detection and peak selection module is used to detect possible target points using the CFAR algorithm, identify possible target signal points and perform peak detection to obtain CFAR detection points, thereby realizing the detection of target points in a noisy background.

[0279] In an exemplary embodiment, the first acquisition module may include: an azimuth DBF&DoA (with Interpolation) module and an elevation DBF&DoA (with Interpolation) module, wherein:

[0280] Azimuth digital beamforming and direction estimation module, used to perform azimuth DBF and DoA estimation on CFAR detection points to obtain the azimuth angle of possible targets;

[0281] The pitch-dimensional digital beamforming and direction estimation module is used to perform pitch-dimensional DBF and DoA estimation on the CFAR detection point to obtain the pitch angle of the possible target; calculate the position information and motion state (speed v) of the possible target in space based on the azimuth and pitch angle of the possible target, as well as the target distance and speed information detected by CFAR; convert the distance, azimuth, pitch angle and speed information of the possible target in the polar coordinate system into the four-dimensional information (x, y, z, v) of the target in the Cartesian coordinate system; and transmit the obtained position and motion state of the possible target in space to the controller.

[0282] In an exemplary embodiment, the second acquisition module may include: a clustering module and a tracking module; wherein,

[0283] A clustering module is used to read the data of CFAR detection points from the MCU and perform cluster analysis on the original point cloud to group the CRAF detection points (i.e., possible targets) into target point clusters;

[0284] The tracking module is used to track the target and generate the target point cloud according to the clustering results.

[0285] The radar data processing device provided in the embodiment of the present application uses a multi-frame joint processing technology, adopts a new inter-frame accumulation method, and combines the detected point cloud clustering and track processing methods to effectively filter out false alarm points and multipath target points, thereby improving the quality of the detected point cloud, thereby providing a guarantee for improving the detection, measurement or recognition performance, and further effectively improving the detection rate, reducing the number of false alarm targets, and thus improving the accuracy of target detection.

[0286] An embodiment of the present application also provides a target detection device, including the radar data processing device described in any one of the embodiments of the present application, and further including: a detection module, which is used to detect the target within the detection area using the target point cloud.

[0287] In an exemplary embodiment, the detection module ( Fig.19 (not shown) can be used to: divide the entire detection area into several preset areas; count the target points detected in each frame according to the preset areas; determine whether there is a target in the detection area and the specific area where it is located based on the number of times the target appears in the preset area.

[0288] The target detection device provided in the embodiment of the present application is particularly suitable for in-cabin human target detection and positioning scenarios with micro-motion and low speed.

[0289] The target detection device provided by the embodiment of the present application effectively filters out false alarm points and multipath target points, thereby realizing a solution for target detection and identification in spatial areas such as automobile cabins, indoors, and factory buildings. While effectively improving the detection rate and reducing the number of false alarm targets, the accuracy of angle estimation and other effects are greatly improved, thereby realizing accurate detection of special targets or weak targets such as infants and young children in the cabin, and realizing applications such as CPD and SBR. Specifically, performance evaluation is carried out through a large number of actual measurement experiments, and the results show that the target detection method provided by the embodiment of the present application effectively filters out false alarm points and multipath points, effectively improves the detection rate, reduces the number of false alarm targets, and thus improves the accuracy of target detection.

[0290] Although the embodiments disclosed in this application are as above, the contents described are only embodiments adopted to facilitate understanding of this application and are not intended to limit this application. Any technician in the field to which this application belongs can make any modifications and changes in the form and details of implementation without departing from the spirit and scope disclosed in this application, but the scope of patent protection of this application shall still be based on the scope defined in the attached claims.

Claims

1. A target detection method, characterized in that: include: Performing cluster analysis on the original point cloud, which includes information reflecting physiological signals; According to the clustering results and the predicted value of the current frame, the target is tracked and the target point cloud is generated; Among them, the prediction value of the current frame is obtained based on the information of the previous frame in the track.

2. The target detection method according to claim 1, further comprising: Counting the number of points in a preset area using the target point cloud; Detect the target in the detection area according to the number of times the target appears in the preset area.

3. The target detection method according to claim 1 or 2, wherein: The original point cloud is a point cloud obtained after processing based on the inter-frame fast Fourier transform FFT.

4. The target detection method according to claim 1 or 2, wherein: The physiological signal includes: a breathing signal and / or a heartbeat signal.

5. The target detection method according to claim 1 or 2, wherein: The possible targets in the original point cloud are grouped into target point clusters using a density-based noise application spatial clustering DBSCAN algorithm, a partitioning clustering algorithm k-means, a density-based algorithm OPTICS, or a hierarchical clustering algorithm Agglomerative.

6. The target detection method according to claim 5, wherein: The original point cloud includes one or any combination of information of the distance, azimuth and elevation angle of the target.

7. The target detection method according to claim 5 or 6, wherein the original point cloud further comprises: speed.

8. The target detection method according to claim 1 or 2, wherein: The clustering analysis of the original point cloud includes: Filter out points that meet the core object requirements from the possible targets in the original point cloud, so that there are at least MinPts points of the core object in the ε neighborhood; The core points in each other's ε neighborhood are connected to form a point cluster, and the points in the ε neighborhood of the core point of the cluster are added to the point cluster to form multiple point clusters, and the center point of each cluster point cloud is used as the cluster center point.

9. The target detection method according to claim 1 or 2, wherein: The predicted value of the current frame is obtained based on the information of the previous frame in the track, including: using the trajectory of the previous frame in each track, or the average of multiple previous frames, or the median of the track points of the past M frames or some characteristic value in a specified order.

10. The target detection method according to claim 1 or 2, wherein: Taking the trajectory of the previous frame in each track as the predicted value of the current frame; The target tracking and target point cloud generation are performed according to the clustering result and the prediction value of the current frame, including: Calculate the distance between the predicted value of each track and all the cluster center points of the current frame respectively, add the cluster center point associated with the current track to the corresponding track according to the calculated distance, and update the track that has matched the cluster center point; The target point cloud is generated according to the original point cloud of a single cluster corresponding to each newly added cluster center point in all tracks.

11. The target detection method according to claim 10, wherein: The step of adding the cluster center point associated with the current track to the corresponding track according to the calculated distance includes: For the cluster center point whose calculated distance is less than the preset threshold, it is considered to match the track, and the cluster center point is added to the corresponding track; For cluster center points whose calculated distance is greater than or equal to the preset threshold, they are added to the candidate tracks for track start determination.

12. The target detection method according to claim 11, wherein: The distance is: Euclidean distance, Chebyshev distance, or Manhattan distance.

13. The target detection method according to claim 10, wherein: The updating of the track that has matched the cluster center point may include: For existing tracks, if no matching cluster center point is found in the current frame, the end flag value of these tracks is increased; For the tracks that are not matched to the cluster center point, check whether the value of their end flag exceeds the preset threshold. If so, remove these tracks; The candidate tracks that meet the MN judgment are added to the formal tracks. The MN judgment criterion is: if the candidate track has N frames that can be associated with the cluster center within the past M frames, it means that the judgment is successful.

14. The target detection method according to claim 10, wherein: The step of generating a target point cloud according to the original point cloud of a single cluster corresponding to each newly added cluster center point in all the tracks includes: The newly added cluster center points in all the tracks are extracted, and the original point cloud and track number of the single cluster corresponding to each cluster center point are output to generate the target point cloud.

15. The target detection method according to claim 2, wherein: The method of using the target point cloud to count the number of points in the preset area includes: Dividing the entire detection area into a plurality of the preset areas; The target points detected in each frame are counted respectively according to the preset area.

16. The target detection method according to claim 2, wherein: The detecting of the target in the detection area comprises: Determine whether there is a target in the detection area.

17. The target detection method according to claim 16, further comprising: Determine a specific area where the target existing in the detection area is located.

18. The target detection method according to any one of claims 1 to 17, before performing cluster analysis on the original point cloud, further comprising: Filtering out the first DC component from the analog-to-digital conversion ADC data; After the data with the first DC component filtered out is windowed, a one-dimensional fast Fourier transform is performed to obtain distance dimension data and store the data; Returning to the step of filtering out the first DC component on the ADC data, until the accumulated distance dimension data reaches a preset number of frames; filtering out a second direct current component from the distance dimension data of the preset number of frames; After the data with the second DC component filtered out is windowed, the range Doppler data is obtained by two-dimensional fast Fourier transform; Incoherent accumulation is performed based on the obtained range-dimensional Doppler data, and constant false alarm rate (CFAR) processing is performed after estimating the noise floor of each range unit of the incoherently accumulated data; as well as The position and motion state of the possible target in space are determined according to the CFAR detection points obtained after CFAR processing to obtain the original point cloud.

19. The radar data processing method according to claim 18, wherein: The preset frame number is 256 frames, 128 frames, 56 frames, or 32 frames.

20. A computer-readable storage medium storing computer-executable instructions, wherein the computer-executable instructions are used to execute the target detection method according to any one of claims 1 to 19.

21. A computer device comprising a memory and a processor, wherein: The memory stores the following instructions executable by the processor: used to execute the steps of the target detection method described in any one of claims 1-19.

22. An electromagnetic wave sensor, characterized in that: include: A signal transmitting module, configured to transmit electromagnetic waves for target detection; A signal receiving module, configured to receive an echo formed by reflection and / or scattering of the electromagnetic wave; as well as A processing module, configured to perform signal and data processing on the echo to achieve determination, positioning and / or identification of the target; Wherein, the processing module is also configured to implement the target detection method as described in claims 1-19.

23. An integrated circuit, characterized in that: It includes a radio frequency module, an analog signal processing module and a digital signal processing module connected in sequence; The radio frequency module is used to generate electromagnetic wave signals and receive echo signals; the analog signal processing module is used to perform frequency reduction processing on the echo signals to obtain intermediate frequency signals; the digital processing module is used to perform analog-to-digital conversion on the intermediate frequency signals to obtain digital signals; as well as The digital processing module performs target judgment, positioning and / or recognition operations based on the target detection method described in any one of claims 1-19.

24. The integrated circuit of claim 23, wherein: The integrated circuit is a millimeter wave chip or a sensor chip.

25. An electromagnetic wave sensor, characterized in that: include: Carrier; An integrated circuit as claimed in claim 23 or 24, arranged on a carrier; An antenna is arranged on the carrier, or the antenna and the integrated circuit are integrated into one device and arranged on the carrier; Wherein, the integrated circuit is connected to the antenna and is used for transmitting the electromagnetic wave signal and / or receiving the echo signal.

26. A terminal device, characterized in that: include: A device body; and an electromagnetic wave sensor as claimed in claim 25 arranged on the device body; wherein the electromagnetic wave sensor is used for target detection and / or communication to provide reference information for the operation of the device body.

Images