Real-time track signal filtering and characteristic continuous track judgment method and system

Abstract

The invention provides a real-time track signal filtering and characteristic continuous track judgment method and system, and belongs to the technical field of millimeter-wave radars. An upper computer receives signal points sent by a millimeter-wave radar, obtains a detection area configured by the upper computer, and screens out signal points outside the detection area to obtain effective signal points; extracting effective signal points in a configuration time period based on a sliding window, and performing clustering analysis on the effective signal points to obtain a coordinate point set; the detection plane of the block detection area is used to obtain a grid set, the coordinate point set is updated to the grid set, the active blocks are associated to generate a trajectory, and the trajectory is subjected to three-dimensional fitting to obtain a continuous trajectory; the characteristic indexes of the continuous tracks are calculated to obtain the effective track characteristics, and the type detection data of the target object is determined according to the judgment result of the effective track characteristics and the configuration characteristics, so that the accuracy of target detection can be improved.

Description

Technical Field

[0001] The present invention relates to millimeter wave radar technology, and in particular to a real-time trajectory signal filtering and feature continuous trajectory judgment method and system. Background Art

[0002] Millimeter wave radar is a wireless sensing technology. It extracts and discovers the location, shape, motion characteristics and other characteristics of the target by analyzing the radar echo characteristics of the received target, thereby determining the number of objects in the spatial area and the location of the objects, which is of great significance for the detection and judgment of the target.

[0003] Most of the existing technologies use the point cloud obtained by the millimeter wave radar to judge the target's movement trajectory. For example, patent document CN118279353A discloses a pedestrian trajectory prediction method and system based on millimeter wave radar, which extracts the pedestrian's historical trajectory signal from the original human body point cloud captured by the millimeter wave radar, and then inputs it into the trajectory prediction model to obtain the future predicted trajectory. However, with the change of the distance of the target relative to the radar and factors such as signal interference, for example, clutter sources such as the ground and buildings in the radar working environment may generate strong reflection signals. These signals interfere with the radar transmission signal to form clutter signals, which reduce the accuracy of target trajectory acquisition, thereby reducing the accuracy of target detection.

[0004] Therefore, how to improve the accuracy of target detection has become an urgent problem that needs to be solved today. Summary of the invention

[0005] The present invention provides a real-time trajectory signal filtering and feature continuous trajectory judgment method and system, which can improve the accuracy of target detection.

[0006] In a first aspect, a real-time trajectory signal filtering and feature continuous trajectory determination method is provided, comprising: The host computer receives the signal points sent by the millimeter wave radar, obtains the detection area configured by the host computer, and filters out the signal points outside the detection area to obtain valid signal points; Extracting valid signal points within a configured time period based on a sliding window, performing cluster analysis on the valid signal points to obtain a set of coordinate points, and generating trajectory data corresponding to the set of coordinate points; Blocking the detection plane of the detection area to obtain a grid collection, updating the trajectory data to the grid collection, associating active blocks to generate a trajectory line, and three-dimensionally fitting the trajectory line to obtain a continuous trajectory; The effective trajectory features of the continuous trajectory are calculated to obtain a judgment index, and the identification detection data of the target object is determined according to the judgment index.

[0007] Optionally, in a possible implementation manner of the first aspect, extracting valid signal points within a configured time period based on a sliding window, clustering and analyzing the valid signal points to obtain a coordinate point set, and generating trajectory data corresponding to the coordinate point set, including: Arrange the radar coordinate points corresponding to each of the valid signal points according to the timestamp to obtain a timestamp sequence, and extract the radar coordinate points within a configured time period in the timestamp sequence based on the sliding window; Based on the timestamp and the extracted radar coordinate points, three-dimensional point data are formed, and cluster analysis is performed on the three-dimensional point data according to the DBSCAN algorithm to obtain multiple coordinate point sets, and trajectory data corresponding to the coordinate point sets are generated.

[0008] Optionally, in a possible implementation manner of the first aspect, segmenting the detection plane of the detection area to obtain a grid collection, updating the trajectory data to the grid collection, associating active blocks to generate a trajectory line, and three-dimensionally fitting the trajectory line to obtain a continuous trajectory, including: Acquire a detection plane corresponding to the detection area, initialize the detection plane according to a preset resolution to obtain a grid collection, and convert the continuous grid collection into a graph data structure; Updating the real-time signal points corresponding to the trajectory data to the grid collection to obtain a signal frame, obtaining the block status of each block in the signal frame, and determining the block in the active state as the active block; A plurality of the signal frames are superimposed, the active blocks are associated to obtain a trajectory line, a three-dimensional coordinate system is constructed, a set of associated signal coordinate points corresponding to the trajectory line is projected to the three-dimensional coordinate system, and continuous coordinate points are fitted to obtain a continuous trajectory.

[0009] Optionally, in a possible implementation manner of the first aspect, obtaining a detection plane corresponding to the detection area, initializing the detection plane according to a preset resolution to obtain a grid collection, and converting the continuous grid collection into a graph data structure, includes: Segmenting the detection plane according to a preset resolution to obtain a grid collection, and determining the block coordinates corresponding to each block in the grid collection; Traversing the grid collection, calculating the coordinates of the adjacent blocks, and saving the block coordinates to the data structure of the corresponding block; Constructing a node set corresponding to each of the blocks, traversing the data structure of each of the blocks, constructing edges of each adjacent block according to the data structure, and adding the edges to the edge set; A graph data structure corresponding to the detection plane is constructed according to the node set and the edge set.

[0010] Optionally, in a possible implementation manner of the first aspect, obtaining a detection plane corresponding to the detection area, initializing the detection plane according to a preset resolution to obtain a grid collection, and converting the continuous grid collection into a graph data structure, includes: Segmenting the detection plane according to a preset resolution to obtain a grid collection, and converting the image frame corresponding to the grid collection into a binary image; Continuous blocks in the binary image are filtered according to the spatial domain pixel algorithm, nodes corresponding to each of the continuous blocks are constructed, edges corresponding to each node are constructed according to the adjacency relationship of each of the continuous blocks, and the graph data structure is generated based on the nodes and edges.

[0011] Optionally, in a possible implementation manner of the first aspect, updating the real-time signal point corresponding to the trajectory data to the grid collection to obtain a signal frame, acquiring a block state of each block in the signal frame, and determining a block in an active state as an active block includes: Acquire the real-time signal point corresponding to the trajectory data, and update the real-time signal point to the grid collection according to the coordinate value to obtain a signal frame; Determine the block activity value of each of the blocks according to the calculation parameters corresponding to each of the blocks in the signal frame, wherein the calculation parameters include the current signal quantity, the current timestamp, the latest signal timestamp of the block and the number of adjacent active blocks; The block state whose block activity value is less than a preset threshold is determined to be an inactive state, the block state whose block activity value is greater than the preset threshold is determined to be an active state, and the block in the active state is determined to be an active block.

[0012] Optionally, in a possible implementation manner of the first aspect, determining a block activity value of each of the blocks according to a calculation parameter corresponding to each of the blocks in the signal frame includes: Counting the current signal quantity of each of the blocks in the signal frame, obtaining the current timestamp of each of the blocks and the block latest signal timestamp of each of the blocks, and determining the number of adjacent active blocks of each of the blocks; Calculate the block activity value of each of the blocks according to the current signal quantity, the current timestamp, the latest signal timestamp of the block and the number of adjacent active blocks; The calculation formula of the block activity value is as follows: in, is the block activity value, is the current number of signals, is the quantity weight, is the current timestamp, The latest signal timestamp of the block, is the time weight, is the number of adjacent active blocks, is the weight of the number of adjacent blocks.

[0013] Optionally, in a possible implementation manner of the first aspect, superimposing a plurality of the signal frames, associating the active blocks to obtain a trajectory line, constructing a three-dimensional coordinate system, projecting a set of associated signal coordinate points corresponding to the trajectory line to the three-dimensional coordinate system, and fitting continuous coordinate points to obtain a continuous trajectory, including: Superimposing the signal frame of a later frame onto the signal frame of a previous frame according to the time sequence, obtaining the active area in the superimposed signal frame, and associating the active blocks to obtain a trajectory line; The active area includes the active area in each signal frame at the historical moment and the active area in the signal frame at the current moment; Generate the three-dimensional coordinate system according to the horizontal coordinate system and the time axis, and project the associated signal coordinate point set onto the time axis and the horizontal coordinate plane; Traversing a plurality of projection points in the projection line, and determining a time interval between each projection point and a previous projection point; The projection points whose time intervals are not within the threshold interval are determined as segmentation points, the projection line is segmented according to the segmentation points to obtain a plurality of continuous segments, and the continuous segments are fitted to obtain a continuous trajectory.

[0014] Optionally, in a possible implementation manner of the first aspect, calculating the effective trajectory feature of the continuous trajectory to obtain a judgment index, and determining the identification detection data of the target object according to the judgment index includes: Acquire valid trajectory features of the continuous trajectory, wherein the valid trajectory features include a timestamp and x, y coordinates; Calculating the effective trajectory features to obtain judgment indicators, the judgment indicators include average speed, instantaneous speed, movement vector, x, y coordinate slope, x, y covariance; The identification track of the target object is determined according to the judgment index, the identification track includes a train track and a pedestrian track, and the identification detection data is obtained according to the identification track.

[0015] In a second aspect, a real-time trajectory signal filtering and feature continuous trajectory judgment system is provided, comprising: A receiving module is used for the host computer to receive the signal points sent by the millimeter wave radar, obtain the detection area configured by the host computer, and filter out the signal points outside the detection area to obtain valid signal points; A clustering module, used to extract valid signal points within a configured time period based on a sliding window, cluster the valid signal points to obtain a set of coordinate points, and generate trajectory data corresponding to the set of coordinate points; A fitting module, used for partitioning the detection plane of the detection area to obtain a grid collection, updating the trajectory data to the grid collection, associating active blocks to generate a trajectory line, and three-dimensionally fitting the trajectory line to obtain a continuous trajectory; The judgment module is used to calculate the effective trajectory features of the continuous trajectory to obtain a judgment index, and determine the recognition detection data of the target object according to the judgment index.

[0016] The beneficial effects of the present invention are as follows: 1. The present invention can convert the plane space into a grid divided by resolution into finite element calculations, merge the coordinate values ​​in a single grid to achieve large data volume coordinate point sampling, reduce the amount of data processing, and improve the efficiency of data processing.

[0017] 2. The present invention can sort continuous grids by signal timestamps and calculate block activity values ​​to realize a block directed graph data structure, which is convenient for calculating the longest path, screening effective feature continuous trajectory signals, and improving the efficiency and accuracy of data screening.

[0018] 3. The present invention can improve the accuracy of recognition by projecting the signal in a single continuous trajectory line on the time axis and secondarily judging the intersection of multiple trajectories and the return of a single trajectory.

[0019] 4. The present invention can detect target movement characteristics (moving speed, moving direction, signal coordinate value variance) and screen the signals that meet the characteristics again to further improve the recognition accuracy.

[0020] 5. The present invention can realize real-time analysis of signals through iterative calculation of current time frame data and historical time frame data, and at the same time, the calculation time complexity of each frame of data is basically the same. BRIEF DESCRIPTION OF THE DRAWINGS

[0021] Figure 1 It is a flow chart of a method for real-time trajectory signal filtering and feature continuous trajectory determination provided by an embodiment of the present invention; Figure 2 is a schematic diagram of a detection plane segmentation and graph data structure conversion provided by an embodiment of the present invention; Figure 3 is a schematic diagram of trajectory line judgment provided by an embodiment of the present invention; Figure 4 It is a structural schematic diagram of a real-time trajectory signal filtering and feature continuous trajectory judgment system provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0022] In order to make the purpose, technical solution and advantages of the embodiments of the present invention clearer, the technical solution in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0023] See also Figure 1 , is a flow chart of a method for real-time trajectory signal filtering and feature continuous trajectory determination provided by an embodiment of the present invention, Figure 1 The execution subject of the method shown may be a software and / or hardware device. The execution subject of the present application may include but is not limited to at least one of the following: user equipment, network equipment, etc. Among them, user equipment may include but is not limited to computers, smart phones, personal digital assistants (PDA) and the electronic devices mentioned above. Network equipment may include but is not limited to a single network server, a server group consisting of multiple network servers, or a cloud based on cloud computing consisting of a large number of computers or network servers, wherein cloud computing is a type of distributed computing, a super virtual computer composed of a group of loosely coupled computers. This embodiment does not limit this. It includes steps S1 to S4, as follows: S1, the host computer receives the signal points sent by the millimeter wave radar, obtains the detection area configured by the host computer, and filters out the signal points outside the detection area to obtain valid signal points.

[0024] In actual applications, there are many stray signals at the railway site, such as vegetation, obstacles, poles, etc., which may cause the signals obtained by the millimeter wave radar to be scattered, reducing the accuracy of judging the target object such as a train. Therefore, when judging the target object, this embodiment will analyze the positional relationship of the coordinate signal over a period of time, screen the valid signal, and fit the continuous motion trajectory to improve the accuracy of target analysis.

[0025] Among them, the host computer refers to a computer device at a higher level in the data acquisition and control system with strong data processing capabilities and control functions. It exchanges data and transmits instructions with the lower computer (such as PLC, single-chip microcomputer or other intelligent devices) through communication interfaces (such as serial ports, network ports, etc.), realizing the functions of monitoring, configuration, data collection and analysis of the lower computer.

[0026] In some embodiments, the detection area configured by the host computer can be manually drawn by the user or can be a preset area shape and size. The user can configure the detection area in conjunction with the drop-down menu bar or list. The shape of the detection area can be a polygon.

[0027] After the host computer receives the signal points sent by the millimeter-wave radar, this embodiment will filter out the data that is not in the detection area, so as to obtain the signal points located in the detection area, that is, the valid signal points. When filtering out the signal points outside the detection area, this embodiment will use the position relationship algorithm between points and polygons for judgment. Among them, the position relationship algorithm judgment of points and polygons is a basic problem in computational geometry, which is used to determine whether a point is located inside, outside or on the boundary of a polygon. This judgment is usually based on geometric concepts such as vectors, rays or angles.

[0028] Specifically, when the host computer receives a new signal point, the host computer can use the point-polygon position relationship algorithm to determine whether the signal point is located in the detection area. If the signal point is outside the detection area, the signal point is filtered out.

[0029] Through step S1, irrelevant data can be removed to improve data processing efficiency, and interference data can also be removed to improve accuracy during data processing.

[0030] S2, extracting valid signal points within a configured time period based on a sliding window, performing cluster analysis on the valid signal points to obtain a coordinate point set, and generating trajectory data corresponding to the coordinate point set.

[0031] When extracting valid signal points within a configured time period through a sliding window, this embodiment first sorts the radar coordinate points corresponding to each valid signal point according to the timestamp, stores data of a certain length of time through a time sequence queue, and then extracts data within the configured time period through a sliding window. In actual applications, the configured time period can be a pre-configured time period.

[0032] For example, assuming that the size of the time sliding window is n, the time data from t to t+n can be retrieved in sequence, where t refers to the current time.

[0033] When clustering the valid signal points, this implementation will combine the read radar coordinate points and timestamps into three-dimensional data, perform cluster analysis on the three-dimensional point data set through the DBSCAN algorithm, obtain the coordinate point set, and then generate the trajectory data corresponding to the coordinate point set.

[0034] Based on the above embodiment, the specific implementation of step S2 may be: The radar coordinate points corresponding to the valid signal points are arranged according to the timestamps to obtain a timestamp sequence, and the radar coordinate points within a configured time period in the timestamp sequence are extracted based on the sliding window.

[0035] The timestamp sequence refers to a sequence obtained by arranging the radar coordinate points according to the timestamp.

[0036] Based on the timestamp and the extracted radar coordinate points, three-dimensional point data are formed, and cluster analysis is performed on the three-dimensional point data according to the DBSCAN algorithm to obtain multiple coordinate point sets, and trajectory data corresponding to the coordinate point sets are generated.

[0037] In practical applications, three-dimensional point data can be (x, y, t) coordinate data.

[0038] Among them, the DBSCAN clustering algorithm is a density-based clustering algorithm, and its full name is "Density-Based Spatial Clustering of Applications with Noise". The DBSCAN clustering algorithm can discover categories of arbitrary shapes and has strong robustness to noise data. The basic idea is that if a point has enough neighboring points in a given neighborhood, then the point is a core point; if a point does not have enough neighboring points in a given neighborhood, then the point is a boundary point; if a point is neither a core point nor a boundary point, then the point is a noise point.

[0039] After obtaining one or more sets of coordinate point sets through the DBSCAN clustering algorithm, the coordinate point data in each set of coordinate point sets can be analyzed to generate trajectory data. In some embodiments, the coordinate point sets can be sorted by timestamp to generate the trajectory.

[0040] Through the above embodiments, coordinate points can be quickly clustered to reduce redundancy and duplication in data, thereby improving the efficiency and accuracy of data processing.

[0041] S3, partitioning the detection plane of the detection area to obtain a grid collection, updating the coordinate point set to the grid collection, associating active blocks to generate a trajectory line, and three-dimensionally fitting the trajectory line to obtain a continuous trajectory.

[0042] After obtaining the trajectory data, this embodiment further determines whether the trajectory is continuous and screens out trajectories that do not meet the requirements, thereby improving the accuracy of target judgment.

[0043] Specifically, when judging whether the trajectory is continuous, the present embodiment will block the detection plane corresponding to the detection area to obtain a grid collection, and quickly judge whether there are coordinate points in adjacent blocks through the coordinate points located in the grid collection, and filter out discrete signal coordinate points to judge whether the remaining coordinate points meet the continuity standard, so that the coordinate values ​​in a single grid can be merged and calculated and large-scale coordinate point sampling can be realized, thereby reducing the complexity of the algorithm and improving the efficiency of data processing.

[0044] Among them, the detection plane refers to the plane space corresponding to the detection area, the grid collection refers to the grid block collection obtained after the detection plane is blocked, updating the coordinate point set to the grid collection refers to distributing the coordinate values ​​corresponding to the coordinate point set to the grid collection, the active block refers to the block that meets the activity level, and associating active blocks helps to accurately express the active connections and dependencies between blocks. The trajectory line refers to the line segment obtained by associating the coordinate points in the active block, and the continuous trajectory refers to the trajectory obtained after fitting the trajectory line in the three-dimensional coordinate system.

[0045] It is understandable that the reason for judging the continuity of the trajectory through a collection of grids is to reduce the complexity of the calculation, so that there is no need to calculate a single coordinate point, but to distribute the coordinate points to the grid, calculate a single grid, and achieve a large number of coordinate point sampling, thereby improving the efficiency of data processing.

[0046] In some embodiments, based on the above embodiments, the specific implementation of step S3 may be: S31, obtaining a detection plane corresponding to the detection area, initializing the detection plane according to a preset resolution to obtain a grid collection, and converting the grid collection into a graph data structure.

[0047] The preset resolution refers to a pre-configured grid size when dividing the detection plane into grids. Specifically, the detection plane can be initialized, and then the detection plane is divided by the preset resolution to obtain a grid collection.

[0048] In order to facilitate the calculation of the longest path and path planning, this embodiment also converts the grid collection into a graph data structure, that is, converts the continuous plane into a graph data structure. Specifically, the information of each block and its adjacent blocks can be determined, nodes are generated through blocks, and the adjacent relationships between blocks form edges. The graph data structure is constructed through nodes and edges.

[0049] See also Figure 2 , is a schematic diagram of a detection plane segmentation and graph data structure conversion provided by an embodiment of the present invention. Figure 2As shown, the left side is the grid collection obtained after the detection plane is segmented, the middle is the graph data structure obtained by converting A1, A2, A6, A8, and A9 in the grid collection, and the right side is the traversal order of the corresponding blocks of A1, A2, A6, A8, and A9. Specifically, when converting the blocks in the grid collection into a graph data structure, each block in the grid collection can be traversed. For example, the traversal order can be A1-A2-A6-A9-A8. After the traversal, the adjacent relationship between each block can be obtained. In this embodiment, the maximum number of adjacent grids is 8, so the adjacent relationships that can be obtained are A1 and A2, A2 and A6, A6 and A9, A9 and A8, and A8 and A6. A1, A2, A6, A8, and A9 are each regarded as a node, and then the edges corresponding to the corresponding nodes are generated according to the adjacent relationships between the blocks. As shown in the figure, the edges corresponding to the nodes A1 and A2, A2 and A6, A6 and A9, A9 and A8, and A8 and A6 can be generated respectively, thereby converting continuous blocks into a graph data structure.

[0050] It is worth mentioning that Figure 2 A schematic diagram of generating a graph data structure only for some blocks. In actual applications, the graph data structure corresponding to all blocks can be generated in the above manner, so that efficient path planning can be performed through the graph data structure.

[0051] In some embodiments, step S31 may be implemented in the following manner: The detection plane is segmented according to a preset resolution to obtain a grid collection, and the block coordinates corresponding to each block in the grid collection are determined.

[0052] Specifically, after the detection plane is segmented to obtain a grid collection, each block in the grid collection may be configured with a block coordinate that uniquely corresponds to it, for example, the coordinates of the center point of the corresponding block.

[0053] The grid collection is traversed, the coordinates of the adjacent blocks are calculated, and the coordinates of the blocks are saved in the data structure of the corresponding blocks.

[0054] Specifically, each block in the grid collection can be accessed in turn. For each block, the coordinates of its adjacent blocks are calculated. Specifically, the coordinates of the blocks above, below, left, right, upper left, lower left, upper right, and lower right can be accessed. The calculated adjacent block coordinates are saved in the data structure of the current block. The data structure can be a list, array, or other structure for storing the adjacent block coordinates, so that each block contains the coordinate information of its adjacent blocks.

[0055] Construct a node set corresponding to each of the blocks, traverse the data structure of each of the blocks, construct edges of each adjacent block according to the data structure, and add the edges to the edge set.

[0056] A node set refers to a set composed of nodes corresponding to all blocks. Combined with the data structure, the adjacent relationship between blocks can be quickly obtained, so that the edges between adjacent blocks can be constructed to obtain an edge set.

[0057] A graph data structure corresponding to the detection plane is constructed according to the node set and the edge set.

[0058] After obtaining the node set and edge set, the continuous blocks can be quickly converted into a graph data structure.

[0059] In some other embodiments, step S31 may also be implemented in the following manner: The detection plane is segmented according to a preset resolution to obtain a grid collection, and the image frame corresponding to the grid collection is converted into a binary image.

[0060] In this embodiment, when converting continuous blocks into a graph data structure, the graph data structure is converted by filtering with a spatial pixel algorithm. Specifically, after obtaining a grid collection, the plane image frame corresponding to the grid collection can be preprocessed, that is, binarized to obtain a binary image, and then the spatial pixel algorithm is applied to the binary image to convert the graph data structure.

[0061] Continuous blocks in the binary image are filtered according to the spatial domain pixel algorithm, nodes corresponding to each of the continuous blocks are constructed, edges corresponding to each node are constructed according to the adjacency relationship of each of the continuous blocks, and the graph data structure is generated based on the nodes and edges.

[0062] The spatial pixel algorithm mainly involves the spatial domain enhancement technology in image processing, which is an enhancement technology that directly acts on image pixels. The spatial pixel algorithm can filter out continuous adjacent blocks in the binary image, and then generate nodes and edges corresponding to each block based on the adjacent relationship between adjacent blocks, thereby converting the grid collection into a graph data structure.

[0063] Through the above method, the adjacent relationship between each block in the grid collection can be quickly determined according to the graph data structure, thereby improving the efficiency of path planning.

[0064] S32, updating the real-time signal points corresponding to the trajectory data to the grid collection to obtain a signal frame, obtaining the block status of each block in the signal frame, and determining the block in the active state as the active block.

[0065] Specifically, the real-time signal points corresponding to the trajectory data can be obtained, and the real-time signal points can be distributed to the grid collection to obtain the image frame after the signal points of the detection plane are updated, that is, the signal frame. The block state is determined by the number of signal coordinates in the block, the coordinate update time, etc., and is divided into active state and inactive state. For the inactive block, the longest path calculation node is removed, so that the signal coordinate point set with correlation of discrete signal output can be filtered out, and the accuracy of subsequent judgment can be improved.

[0066] In some embodiments, step S32 may be implemented by the following steps: The real-time signal points corresponding to the trajectory data are obtained, and the real-time signal points are updated according to the coordinate values ​​to obtain a signal frame in the grid collection.

[0067] When acquiring signal frames, real-time signal points can be first acquired based on the millimeter-wave radar, and then the real-time signal points are distributed to a grid collection according to the coordinate values ​​to acquire an image frame corresponding to the plane after the signal points are updated.

[0068] The block activity value of each of the blocks is determined according to the calculation parameters corresponding to each of the blocks in the signal frame, wherein the calculation parameters include the current signal quantity, the current timestamp, the latest signal timestamp of the block and the number of adjacent active blocks.

[0069] After the signal point is updated, this embodiment will superimpose a double judgment of the block activity value in the spatial domain, thereby filtering the discrete signal.

[0070] Among them, the calculation parameters refer to the parameters used when calculating the block activity value, the block activity value refers to the numerical value indicating the activity level of each block, the current number of signals refers to the number of current signal points in each block, the current timestamp refers to the time point when the block activity value is calculated, the block latest signal timestamp refers to the time point when the latest signal is received in the block, and the number of adjacent active blocks refers to the number of active blocks adjacent to the current block.

[0071] Specifically, the block activity value can be calculated by the following embodiments: The current signal quantity of each of the blocks in the signal frame is counted, the current timestamp of each of the blocks and the block latest signal timestamp of each of the blocks are obtained, and the quantity of adjacent active blocks of each of the blocks is determined.

[0072] The block activity value of each block is calculated according to the current signal quantity, the current timestamp, the latest signal timestamp of the block and the number of adjacent active blocks.

[0073] The calculation formula of the block activity value is as follows: The calculation formula of the block activity value is as follows: in, is the block activity value, The current number of signals, which can be used to measure the activity of the block. The more signals there are, the more active the block is.

[0074] is the quantity weight, used to adjust The degree of influence on the block activity value. The weight can be set according to actual needs to reflect the importance of the number of signals in evaluating the block activity.

[0075] It is the current timestamp, which is the time point when the active value of the current block is calculated. The latest signal timestamp of the block, by comparing and , the time difference from the generation of the signal to the current time can be calculated. This time difference reflects the real-time nature of the signal within the block.

[0076] is the time weight, used to adjust the time difference The degree of influence on the block activity value. The weight can be set according to actual needs to reflect the importance of signal real-time in evaluating block activity.

[0077] is the number of adjacent active blocks. The number of adjacent active blocks can reflect the spatial activity of the current block, because the activity of adjacent blocks may mean that the current block is also active.

[0078] The weight of the number of adjacent blocks, used to adjust The degree of influence on the activity value of the block. The weight can be set according to actual needs to reflect the importance of the activity of adjacent blocks in evaluating the activity of the current block.

[0079] Through the above calculation method, the activity value of the relevant block can be calculated in each frame signal period, thereby improving the accuracy of processing.

[0080] The block state whose block activity value is less than a preset threshold is determined to be an inactive state, the block state whose block activity value is greater than the preset threshold is determined to be an active state, and the block in the active state is determined to be an active block.

[0081] The preset threshold refers to a preset threshold used to determine the activity level of a block. If the value is less than the threshold, the corresponding block can be marked as inactive. If the value is greater than the threshold, the corresponding block can be marked as active and the corresponding block can be determined as an active grid.

[0082] S33, superimposing a plurality of the signal frames, associating the active blocks to obtain a trajectory line, constructing a three-dimensional coordinate system, projecting a set of associated signal coordinate points corresponding to the trajectory line to the three-dimensional coordinate system, and fitting continuous coordinate points to obtain a continuous trajectory.

[0083] After obtaining multiple active blocks, the signal frames at different time points can be superimposed, and the longest path of the associated active blocks can be calculated for the graph data to obtain the trajectory line, and a three-dimensional coordinate system is constructed based on the horizontal coordinate system and the time axis. The set of associated signal coordinate points in the trajectory line is projected onto the time axis and the horizontal coordinate plane. By checking whether the time projection line is continuous or not, the spatial and temporal continuous coordinate points are selected to fit the moving trajectory curve, i.e., the continuous trajectory. Among them, the trajectory line is the curve constructed by the filtered coordinate points, and the associated signal coordinate points are the coordinate points corresponding to the trajectory points.

[0084] In some embodiments, step S33 may be implemented by the following steps: The signal frame of a later frame is superimposed on the signal frame of a previous frame according to the time sequence, an active area in the superimposed signal frame is obtained, and the active block is associated to obtain a trajectory line.

[0085] The active area includes the active area in each signal frame at a historical moment and the active area in the signal frame at a current moment.

[0086] It can be understood that in order to obtain the activity status of signal points within a certain period of time, the signal frames within a certain period of time can be superimposed in chronological order, and then the active areas in the superimposed signal frames can be associated, so that discrete signal coordinate points can be filtered out, wherein the coordinate points within the active block are associated during the association.

[0087] The three-dimensional coordinate system is generated according to the horizontal coordinate system and the time axis, and the associated signal coordinate point set is projected onto the time axis and the horizontal coordinate plane.

[0088] The horizontal coordinate system refers to a two-dimensional plane, which represents the horizontal position of an object. The time axis refers to an axis perpendicular to the horizontal coordinate system, which is used to represent the change of time. Combining the time axis with the horizontal coordinate system forms a three-dimensional coordinate system (x, y, t), where x and y represent horizontal position and t represents time.

[0089] Projection refers to the process of mapping the position of a point or set in a certain direction to another plane or axis. In this embodiment, the associated signal coordinate point set in each trajectory line is projected onto the time axis and the horizontal coordinate plane, that is, the yt plane.

[0090] A plurality of projection points in the projection line are traversed to determine the time interval between each projection point and the previous projection point.

[0091] It is understandable that the reason for traversing multiple projection points in the projection line and determining the time interval between adjacent projection points is to judge the continuity of the trajectory line in time. If the points on the projection line are continuous, it means that the movement of the object during this period of time is continuous. If breakpoints or intervals appear on the projection line, it means that the movement of the object during this period of time may be disturbed or interrupted.

[0092] The projection points whose time intervals are not within the threshold interval are determined as segmentation points, the projection line is segmented according to the segmentation points to obtain a plurality of continuous segments, and the continuous segments are fitted to obtain a continuous trajectory.

[0093] The threshold interval is used to judge the time continuity. If the time interval is not within the threshold interval, it means that the trajectory line may be interrupted. Then the corresponding projection point can be used as a segmentation point. The projection line can be segmented based on the segmentation point to obtain multiple continuous segments, and then the continuous segments can be fitted to obtain a continuous trajectory.

[0094] See also Figure 3 , which is a schematic diagram of trajectory line judgment provided by an embodiment of the present invention. Figure 3 The signal frames in the image are arranged in chronological order, from top to bottom and from left to right. Figure 3 It can be seen that each signal frame is obtained by superimposing the previous signal frame. The pixel value of each coordinate point is related to its corresponding weight. Each coordinate point has its own life cycle and becomes inactive after a period of time, so its corresponding pixel value can change from black to gray. It is worth mentioning that when associating active blocks, the historical active blocks are associated with the current active blocks, so there will be gray coordinate points and black coordinate points when associating.

[0095] like Figure 3 As shown in the last two figures in , after the associated active block obtains the associated signal coordinate point set, it can be projected onto the yt plane to obtain multiple continuous segments. By fitting the continuous segments, multiple continuous trajectories can be obtained.

[0096] Through the above method, discrete signals can be screened out, and trajectory signals with effective continuous features can be quickly screened out, thereby improving the efficiency of data processing and the accuracy of target recognition.

[0097] S4, calculating effective trajectory features of the continuous trajectory to obtain a judgment index, and determining identification and detection data of the target object according to the judgment index.

[0098] After obtaining the continuous trajectory, its feature data can be calculated. Specifically, the timestamp and x, y coordinates of the continuous trajectory can be used to calculate the judgment index, which includes average speed, instantaneous speed, moving vector, x, y coordinate slope, and x, y covariance. Among them, the effective trajectory feature refers to the trajectory data of the continuous trajectory, the judgment index refers to the index when performing target recognition judgment, the target refers to the object to be judged, and the recognition detection data refers to the data after the target is judged and identified.

[0099] Based on the above embodiment, the specific implementation of step S4 may be: Acquire valid trajectory features of the continuous trajectory, where the valid trajectory features include a timestamp and x, y coordinates.

[0100] The effective trajectory features are calculated to obtain judgment indicators, which include average speed, instantaneous speed, movement vector, x, y coordinate slope, and x, y covariance.

[0101] Among them, the average speed can be calculated by dividing the total distance by the total time, the instantaneous speed can be calculated by dividing the distance between adjacent points by the time interval, the moving vector refers to the change between adjacent points, the x, y coordinate slope can be calculated by linear regression, and the x, y covariance can measure the linear relationship between two variables.

[0102] The identification track of the target object is determined according to the judgment index, the identification track includes a train track and a pedestrian track, and the identification detection data is obtained according to the identification track.

[0103] It is understandable that the trajectories of pedestrians and lists are different, so after calculating multiple judgment indicators, the action trajectory of the target object can be identified by the judgment indicators to obtain the identification trajectory. In practical applications, after all features are extracted, these features can be used to train a classifier, such as SVM, decision tree, random forest, etc., to distinguish between train and pedestrian trajectories.

[0104] Through the above method, based on the movement characteristics of the detected target, such as movement speed, movement direction, signal coordinate variance, etc., the characteristic signals can be screened again to further improve the recognition accuracy.

[0105] See also Figure 4 , is a structural diagram of a real-time trajectory signal filtering and feature continuous trajectory judgment system provided by an embodiment of the present invention, the real-time trajectory signal filtering and feature continuous trajectory judgment system comprising: A receiving module is used for the host computer to receive the signal points sent by the millimeter wave radar, obtain the detection area configured by the host computer, and filter out the signal points outside the detection area to obtain valid signal points; A clustering module, used to extract valid signal points within a configured time period based on a sliding window, cluster the valid signal points to obtain a set of coordinate points, and generate trajectory data corresponding to the set of coordinate points; A fitting module, used for partitioning the detection plane of the detection area to obtain a grid collection, updating the trajectory data to the grid collection, associating active blocks to generate a trajectory line, and three-dimensionally fitting the trajectory line to obtain a continuous trajectory; The judgment module is used to calculate the effective trajectory features of the continuous trajectory to obtain a judgment index, and determine the recognition detection data of the target object according to the judgment index.

[0106] Figure 4 The apparatus of the embodiment shown can be used to perform Figure 1 The implementation principles and technical effects of the steps in the method embodiment shown are similar and will not be repeated here.

[0107] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or replace some or all of the technical features therein by equivalents. However, these modifications or replacements do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of the present invention.

Claims

1. A real-time trajectory signal filtering and feature continuous trajectory judgment method, characterized in that: include: The host computer receives the signal points sent by the millimeter wave radar, obtains the detection area configured by the host computer, and filters out the signal points outside the detection area to obtain valid signal points; Extracting valid signal points within a configured time period based on a sliding window, performing cluster analysis on the valid signal points to obtain a set of coordinate points, and generating trajectory data corresponding to the set of coordinate points; Blocking the detection plane of the detection area to obtain a grid collection, updating the trajectory data to the grid collection, associating active blocks to generate a trajectory line, and three-dimensionally fitting the trajectory line to obtain a continuous trajectory; The effective trajectory features of the continuous trajectory are calculated to obtain a judgment index, and the identification detection data of the target object is determined according to the judgment index.

2. The method according to claim 1, characterized in that: Extracting valid signal points within a configured time period based on a sliding window, performing cluster analysis on the valid signal points to obtain a coordinate point set, and generating trajectory data corresponding to the coordinate point set, including: Arrange the radar coordinate points corresponding to each of the valid signal points according to the timestamp to obtain a timestamp sequence, and extract the radar coordinate points within a configured time period in the timestamp sequence based on the sliding window; Based on the timestamp and the extracted radar coordinate points, three-dimensional point data are formed, and cluster analysis is performed on the three-dimensional point data according to the DBSCAN algorithm to obtain multiple coordinate point sets, and trajectory data corresponding to the coordinate point sets are generated.

3. The method according to claim 1, characterized in that Blocking the detection plane of the detection area to obtain a grid collection, updating the trajectory data to the grid collection, associating active blocks to generate trajectory lines, and three-dimensionally fitting the trajectory lines to obtain a continuous trajectory, including: Acquire a detection plane corresponding to the detection area, initialize the detection plane according to a preset resolution to obtain a grid collection, and convert the continuous grid collection into a graph data structure; Updating the real-time signal points corresponding to the trajectory data to the grid collection to obtain a signal frame, obtaining the block status of each block in the signal frame, and determining the block in the active state as the active block; A plurality of the signal frames are superimposed, the active blocks are associated to obtain a trajectory line, a three-dimensional coordinate system is constructed, a set of associated signal coordinate points corresponding to the trajectory line is projected to the three-dimensional coordinate system, and continuous coordinate points are fitted to obtain a continuous trajectory.

4. The method according to claim 3, characterized in that: Acquiring a detection plane corresponding to the detection area, initializing the detection plane according to a preset resolution to obtain a grid collection, and converting the grid collection into a graph data structure, including: Segmenting the detection plane according to a preset resolution to obtain a grid collection, and determining the block coordinates corresponding to each block in the grid collection; Traversing the grid collection, calculating the coordinates of the adjacent blocks, and saving the block coordinates to the data structure of the corresponding block; Constructing a node set corresponding to each of the blocks, traversing the data structure of each of the blocks, constructing edges of each adjacent block according to the data structure, and adding the edges to the edge set; A graph data structure corresponding to the detection plane is constructed according to the node set and the edge set.

5. The method according to claim 3, characterized in that: Acquiring a detection plane corresponding to the detection area, initializing the detection plane according to a preset resolution to obtain a grid collection, and converting the grid collection into a graph data structure, including: Segmenting the detection plane according to a preset resolution to obtain a grid collection, and converting the image frame corresponding to the grid collection into a binary image; Continuous blocks in the binary image are filtered according to the spatial domain pixel algorithm, nodes corresponding to each of the continuous blocks are constructed, edges corresponding to each node are constructed according to the adjacency relationship of each of the continuous blocks, and the graph data structure is generated based on the nodes and edges.

6. The method according to claim 3, characterized in that: Updating the real-time signal points corresponding to the trajectory data to the grid collection to obtain a signal frame, acquiring the block status of each block in the signal frame, and determining the block in the active state as the active block, including: Acquire the real-time signal point corresponding to the trajectory data, and update the real-time signal point to the grid collection according to the coordinate value to obtain a signal frame; Determine the block activity value of each of the blocks according to the calculation parameters corresponding to each of the blocks in the signal frame, wherein the calculation parameters include the current signal quantity, the current timestamp, the latest signal timestamp of the block and the number of adjacent active blocks; The block state whose block activity value is less than a preset threshold is determined to be an inactive state, the block state whose block activity value is greater than the preset threshold is determined to be an active state, and the block in the active state is determined to be an active block.

7. The method according to claim 6, characterized in that Determining a block activity value of each of the blocks according to a calculation parameter corresponding to each of the blocks in the signal frame includes: Counting the current signal quantity of each of the blocks in the signal frame, obtaining the current timestamp of each of the blocks and the block latest signal timestamp of each of the blocks, and determining the number of adjacent active blocks of each of the blocks; Calculate the block activity value of each of the blocks according to the current signal quantity, the current timestamp, the latest signal timestamp of the block and the number of adjacent active blocks; The calculation formula of the block activity value is as follows: in, is the block activity value, is the current number of signals, is the quantity weight, is the current timestamp, The latest signal timestamp of the block, is the time weight, is the number of adjacent active blocks, is the weight of the number of adjacent blocks.

8. The method according to claim 3 or 7, characterized in that: Superimposing a plurality of the signal frames, associating the active blocks to obtain a trajectory line, constructing a three-dimensional coordinate system, projecting a set of associated signal coordinate points corresponding to the trajectory line to the three-dimensional coordinate system, and fitting continuous coordinate points to obtain a continuous trajectory, including: Superimposing the signal frame of a later frame onto the signal frame of a previous frame according to the time sequence, obtaining the active area in the superimposed signal frame, and associating the active blocks to obtain a trajectory line; The active area includes the active area in each signal frame at the historical moment and the active area in the signal frame at the current moment; Generate the three-dimensional coordinate system according to the horizontal coordinate system and the time axis, and project the associated signal coordinate point set onto the time axis and the horizontal coordinate plane; Traversing a plurality of projection points in the projection line, and determining a time interval between each projection point and a previous projection point; The projection points whose time intervals are not within the threshold interval are determined as segmentation points, the projection line is segmented according to the segmentation points to obtain a plurality of continuous segments, and the continuous segments are fitted to obtain a continuous trajectory.

9. The method according to claim 1, characterized in that: Calculating the effective trajectory features of the continuous trajectory to obtain a judgment index, and determining the identification and detection data of the target object according to the judgment index, including: Acquire valid trajectory features of the continuous trajectory, wherein the valid trajectory features include a timestamp and x, y coordinates; Calculating the effective trajectory features to obtain judgment indicators, the judgment indicators include average speed, instantaneous speed, movement vector, x, y coordinate slope, x, y covariance; The identification track of the target object is determined according to the judgment index, the identification track includes a train track and a pedestrian track, and the identification detection data is obtained according to the identification track.

10. A real-time trajectory signal filtering and feature continuous trajectory judgment system, characterized in that: include: A receiving module is used for the host computer to receive the signal points sent by the millimeter wave radar, obtain the detection area configured by the host computer, and filter out the signal points outside the detection area to obtain valid signal points; A clustering module, used to extract valid signal points within a configured time period based on a sliding window, cluster the valid signal points to obtain a set of coordinate points, and generate trajectory data corresponding to the set of coordinate points; A fitting module, used for partitioning the detection plane of the detection area to obtain a grid collection, updating the trajectory data to the grid collection, associating active blocks to generate a trajectory line, and three-dimensionally fitting the trajectory line to obtain a continuous trajectory; The judgment module is used to calculate the effective trajectory features of the continuous trajectory to obtain a judgment index, and determine the recognition detection data of the target object according to the judgment index.

Images