A Track Extrapolation Method for Target Tracking

Abstract

The present invention discloses a track extrapolation method for target tracking. By combining multiple hypotheses on the basis of existing extrapolation, a method of adding a hypothesized target at the position where track loss association occurs is adopted. And when the target still fails to match the point cloud after extrapolating for a certain period of time, the target retreats to the position before the loss of association, significantly reducing the possibility of target tracking errors caused by sudden stillness of the target.

Description

Technical Field

[0001] The present invention relates to the technical field of millimeter-wave radar target tracking, and particularly relates to a track extrapolation method for target tracking. Background Art

[0002] Person detection and quantity statistics are indispensable data in many existing scenarios, such as airports, shopping malls, chain stores, stations, museums, conference rooms, etc. Detecting and counting people and data estimation play important roles in security management and improving the overall service level. Common person tracking and detection technologies mainly use single recognition methods such as images or infrared. Images have certain impacts on privacy and their application scope is limited. In the case of infrared, high-resolution devices are costly, and problems such as being easily interfered are relatively serious. With the progress of technology, the use of millimeter-wave radar sensors can make up for these deficiencies in person counting and has broad application prospects.

[0003] In existing person detection and counting applications, there are mainly camera recognition, traditional manual counting, automatic counting mechanical devices, access control systems using RFID, etc. These methods are restricted by aspects such as cost, accuracy, and implementation difficulties, making it difficult to achieve a balance between economy and accuracy in person detection and counting applications. Existing track extrapolation is basically when the target is lost, maintaining the speed of the track at the previous moment to move. When the target suddenly stops, the target is often extrapolated to a wrong position, resulting in an incorrect judgment of the target position. When the target moves again, a new target is generated at the original position, leading to an incorrect total number of targets. Summary of the Invention

[0004] The object of the present invention is to propose an extrapolation method combining multiple hypotheses and fallback for the problem that in the current situation where millimeter-wave radar is difficult to detect dynamic point clouds of stationary targets, resulting in incorrect tracking positions of suddenly stationary tracks during movement. It measures and tracks the target trajectory in a determined space area to be measured, significantly reducing the tracking error rate of the target in this situation and improving the reliability of target tracking.

[0005] The present invention achieves the above object through the following technical solutions:

[0006] A track extrapolation method for target tracking, comprising the following steps:

[0007] Step 1, emit electromagnetic wave signals to the space range to be measured through a millimeter-wave radar, receive the signals reflected back within the space range to be measured through the millimeter-wave radar, and process the received signals to obtain the point cloud information of the current frame;

[0008] Step 2: Associate the generated track with the point cloud of the current frame. If there is no suitable measurement point cloud associated with the existing track in the current frame, the extrapolation state is maintained.

[0009] Step 3: When a track fails to associate with a suitable measurement point cloud for several consecutive frames, perform multiple hypothesis tracking on the track in extrapolation. Select a position before the track extrapolation based on the association situation before the track extrapolation and perform multiple hypothesis tracking simultaneously.

[0010] Step 4: If neither of the two hypotheses is associated with a suitable measurement point cloud after the track has been extrapolated for several frames, select a position before the track extrapolation based on the association situation before the track extrapolation.

[0011] A further solution is that in Step 1, a millimeter-wave radar is used to emit electromagnetic wave signals to the space range to be measured, and the reflected electromagnetic wave echo signals are analyzed. In order to complete the measurement and tracking of the trajectories of moving targets in the received signals, Fourier transform and signal processing are also applied to obtain the measurements of the radial velocity, position, azimuth angle, and elevation angle of the point cloud targets.

[0012] A further solution is that in Step 2, the clustered point clouds are integrated, and the track and the measurement point cloud are associated based on the distance between the target track and the clustered measurement point cloud, the signal-to-noise ratio (SNR) of the measurement point cloud, and the velocity between the point cloud and the target track. If there is no measurement that meets the conditions near the track, the track in this frame maintains the extrapolation state, that is, the predicted state of the current frame is used as the final state of the track in this frame.

[0013] A further solution is that in Step 3, if the track remains in the extrapolation state for 3 consecutive frames, a new hypothesis is added for the current track for association. The coordinates of the new hypothesis depend on the association situation within the previous 30 frames of the track. If the position (X0, Y0) was measured when the track was associated with more than the point number threshold N of measurement points within the previous 30 frames, then the position (X0, Y0) is used as the position (X H2 , Y H2 ); if the track has not been associated with a measurement that meets the above conditions within 3 s, the measurement position (X1, Y1) where the track was last associated before entering extrapolation is used as the position of the new hypothesis. In the subsequent frames, the two hypotheses are associated simultaneously. During the extrapolation process, after a hypothesis is associated with a measurement point cloud, the hypothesis with the higher association score is selected from the two hypotheses as the only extrapolation position to end the process of extrapolation with multiple hypotheses.

[0014] A further solution is that in Step 4, if neither of the two hypotheses is associated with a suitable measurement point cloud after the track has been extrapolated for 15 frames, according to the above conditions, the target track is also retreated to the position (X0, Y0) or the position (X1, Y1) as the current position (X H1 , Y H1)。

[0015] The beneficial effects of the present invention are as follows:

[0016] A track extrapolation method for target tracking according to the present invention adds a hypothesized target at the position where track loss association occurs by combining multiple hypotheses on the basis of existing extrapolation. And when no point cloud is matched after the target is extrapolated for a certain period of time, the target retreats to the position before the loss of association, significantly reducing the possibility of target tracking errors caused by the target suddenly stopping. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the following will briefly introduce the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are only some embodiments of the present invention. For those of ordinary skill in the art, other drawings can be obtained based on these drawings without creative efforts.

[0018] Figure 1 It is a schematic flowchart of the extrapolation process in the present invention.

[0019] Figure 2 It is a schematic diagram of the point cloud obtained after signal processing of millimeter-wave radar signals. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0020] To make the objectives, technical solutions, and advantages of the present invention clearer, the technical solutions of the present invention will be described in detail below. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all of the embodiments. All other embodiments obtained by those of ordinary skill in the art without creative efforts based on the embodiments of the present invention fall within the scope protected by the present invention.

[0021] In any embodiment, as Figure 1-2 shown, a track extrapolation method for target tracking according to the present invention includes:

[0022] 1. Transmit electromagnetic wave signals through a millimeter-wave radar to the measured space range, receive the signals reflected back within the measured space range by the millimeter-wave radar, and process the received signals to obtain the point cloud information detected by the radar in the current frame.

[0023] 2. Associate the generated track with the point cloud in the current frame. If the existing track does not associate with a suitable measurement point cloud in the current frame, maintain the extrapolation state.

[0024] 3. When the track does not associate with a suitable measurement point cloud for 3 consecutive frames, perform multi-hypothesis tracking on the track entering extrapolation, and select a position before the track extrapolation according to the association situation before the track extrapolation to perform multi-hypothesis tracking simultaneously.

[0025] 4. If neither of the two hypotheses is associated with a suitable measurement point cloud after the track is extrapolated to 15 frames, the position before the track extrapolation is selected for the track according to the association situation before the extrapolation.

[0026] The following is an explanation of these 4 parts respectively:

[0027] (1) Transmit electromagnetic wave signals to the space range to be measured. The millimeter-wave radar uses a 60 GHz millimeter-wave broadband radar to adjust the radar transmission and reception parameters. The space range to be measured can be adjusted from within 10m * 10m to within 100m * 100m, and can be applied to small-range application scenarios such as indoor meeting rooms, or used for outdoor application scenarios such as scenic spots. Perform a fast Fourier transform (FFT) on the collected signals, transform the collected signals from the time domain to the frequency domain, analyze the collected radar signals from the frequency domain characteristics, and obtain a point cloud containing information on distance, azimuth angle, elevation angle, Doppler velocity, and signal-to-noise ratio. The point cloud distribution is as shown by the small dots in Figure 2 , and the circles in Figure 2 represent the target track positions.

[0028] (2) Figure 1 In 101, the generated point cloud is clustered into point stacks P n , and each point stack contains multiple point clouds with relatively close positions. The position of the point stack is the average position of all the point clouds contained in all the point stacks, and the point stack snr is the average snr of all the point clouds contained.

[0029] (3) Figure 1 In 102 - 105, the target track is associated with the measurement point cloud. If there is a point stack of measurement points within a certain distance range of the target track, the track is associated with the point cloud; otherwise, the target track has no association in the current frame. If the track is associated, combined with the Kalman filter, the predicted position and measurement of the current frame of the track are integrated to obtain the final position of the current frame of the track, and then wait for the measurement of the next frame. If the track has no association in the current frame but has been associated within 3 frames, the predicted position of the current frame of the track is used as the final position of the current frame of the track, and then wait for the measurement of the next frame.

[0030] (4) Figure 1 In 106, if when making a judgment in Figure 1 104, the track target has lost association for 3 consecutive frames and the track is updated relying on the predicted position, then multi-hypothesis tracking is adopted. The predicted position of the current frame of the track (X H1 , Y H1As the hypothetical position 1, on this basis, a new position hypothesis is added as the hypothetical position 2, and the new hypothetical coordinates depend on the association situation within the first 30 frames of the track. If the track has been associated with more than the point threshold N of measurement point clusters R within 30 frames i , then the position (X0, Y0) of the point cluster R i is used as the position P2(X H2 , Y H2 ) of the new hypothesis; if the track has not been associated with a measurement that meets the above conditions within 3s, then the position (X1, Y1) of the measurement that the track was last associated with before entering extrapolation is used as the position P1(X H1 , Y H1 ) of the new hypothesis.

[0031] (5) Figure 1 In (5), 107 is one frame after 106. Two hypotheses of the target track are simultaneously associated with the measurements of the current frame. If one of the two hypotheses of the track is associated with a measurement, the multiple hypotheses are cancelled, and the measurement position is used as the final position of the target in the current frame; if both hypotheses are associated with measurement values in the current frame, then it is determined which measurement has a higher score value according to the distance between the track and the associated measurement, the number of points of the associated measurement, and the signal-to-noise ratio of the associated measurement. The measurement with the higher score is used as the associated measurement of the track to update the track position as the final position of the track in the current frame, and the multiple hypotheses are cancelled.

[0032] (6) Figure 1 In (6), 109 and 111 are in step (5). If neither of the two hypotheses of the target track is associated with a measurement in the current frame and the number of consecutive lost measurement frames of the track is less than 15 frames, continue to extrapolate at the predicted speed, and the extrapolated position is used as the hypothesis 1 position; keep the position (X H2 , Y H2 ) as the position of hypothesis 2; then keep the two hypotheses and enter the next frame to wait for the point cloud of the next frame.

[0033] (7) Figure 1 In (7), 110 is in step (6). If the target track has lost measurements for 15 consecutive frames and has not been associated with a measurement, cancel the multiple hypotheses, end the extrapolation, and return to the position P2(X H2 , Y H2 ) before extrapolation. The selection method of P2 is the same as that in step (4), and it is used as the only final position of the track in the current frame.

[0034] As described above, it is only the specific implementation manner of the present invention, but the protection scope of the present invention is not limited thereto. Any person skilled in the art within the technical scope disclosed by the present invention can easily think of changes or substitutions, which should all be covered within the protection scope of the present invention. Therefore, the protection scope of the present invention shall be subject to the protection scope of the claims. Additionally, it should be noted that in the above specific implementation manner, the various specific technical features described can be combined in any appropriate manner without contradiction. To avoid unnecessary repetition, the present invention will not separately describe various possible combination methods. Furthermore, any arbitrary combination can be made between different implementation manners of the present invention, as long as it does not violate the idea of the present invention, it should also be regarded as the content disclosed by the present invention.

Claims

1. A track extrapolation method for target tracking, characterized in that, Including the following steps: Step 1: Transmit electromagnetic wave signals to the space range to be measured through a millimeter-wave radar, receive the signals reflected back within the space range to be measured through the millimeter-wave radar, and process the received signals to obtain the point cloud information of the current frame; Step 2: Associate the generated track with the point cloud of the current frame. If there is no appropriate measurement point cloud associated with the existing track in the current frame, maintain the extrapolated state; in Step 2, comprehensively consider the clustered point clouds, and associate the track and the measurement point cloud according to the distance between the target track and the clustered measurement point cloud, the signal-to-noise ratio of the measurement point cloud, and the speed of the point cloud and the target track; in Step 2, if there is no measurement that meets the conditions near the track, the track of this frame remains in the extrapolated state, that is, the predicted state of the current frame is used as the final state of the track in this frame; Step 3, when a track fails to be associated with a suitable measurement point cloud for several consecutive frames, perform multiple hypothesis tracking on the track entering extrapolation, and select a position before track extrapolation according to the association situation before track extrapolation to simultaneously perform multiple hypothesis tracking; in Step 3, if the track remains in the extrapolation state within 3 consecutive frames, a new hypothesis is added to the current track for association, and the coordinates of the new hypothesis depend on the association situation within the previous 30 frames of the track; if the position (X0, Y0) measured when the track was last associated with more than the point number threshold N of measurement points within 30 frames, then the position (X0, Y0) is used as the position (X H2 , Y H2 ) of the new hypothesis; if the track fails to be associated with a measurement satisfying the above conditions within 3 s, then the measurement position (X1, Y1) associated with the track the last time before entering extrapolation is used as the position of the new hypothesis. In subsequent frames, the two hypotheses are simultaneously associated. During the extrapolation process, after a hypothesis is associated with a measurement point cloud, the hypothesis with a higher association score is selected from the two hypotheses as the only extrapolation position to end the process of extrapolating multiple hypotheses; Step 4: If neither of the two hypotheses is associated with an appropriate measurement point cloud after extrapolating the track for several frames, select a position before the track extrapolation for the track according to the association situation before extrapolation.

2. The track extrapolation method for target tracking according to claim 1, characterized in that, In Step 1, a millimeter-wave radar is used to transmit electromagnetic wave signals to the space range to be measured, and the reflected electromagnetic wave echo signals are analyzed to obtain the measurements of the radial velocity, position, azimuth angle, and elevation angle of the point cloud target.

3. The track extrapolation method for target tracking according to claim 1, characterized in that, If neither of the two hypotheses is still associated with a suitable measurement point cloud after 15 frames of track extrapolation in step 4, the target track is retreated to position (X0, Y0) or position (X1, Y1) as the current position (X H1 , Y H1 )

Images