A filter-controlled method against velocity distortions in the target overlap region during target tracking

Abstract

The invention is related to a filter-controlled method for target tracking in which velocity distortions in the overlap region caused by an obstacle in front of the target sensor are avoided. A filter-controlled method for preventing speed distortions in the overlap region according to the invention comprising the process steps, obtaining measurement data from a sensor (1), clustering the incoming measurement data, generating a plot as a result of the clustering, and determining that the target (2) has entered the semi-overlap region (5) by comparing the plots obtained from the previous step of the target tracking filter, running the forward prediction and update phases of the tracking filter when the target is not in the semi-overlap region (5), and running only the forward prediction phase of the tracking filter when the target enters the overlap region (4).

Description

[0001] A FILTER-CONTROLLED METHOD AGAINST VELOCITY DISTORTIONS IN THE TARGET OVERLAP REGION DURING TARGET TRACKING

[0002] Technical Field

[0003] The invention is related to a filter-controlled method for target tracking in which velocity distortions in the overlap region caused by an obstacle in front of the target sensor are avoided.

[0004] Prior Art

[0005] During target tracking with Lidar and similar sensors, an overlap region appears due to an obstacle in front of the sensor. Normally, during target tracking, a plot is generated by clustering multiple measurements taken by the sensor from different points of the target, and kinematic data such as position and velocity of the target are determined with target tracking filters.

[0006] As the target starts to enter the overlap region, the target will appear to be smaller in size due to the smaller number of measurements taken by the sensor, and therefore the predicted velocity of the target will decrease. Similarly, as the target exits the overlap region, a lower velocity will be predicted and the target will appear to be accelerating. This region of velocity degradation can be referred to as the semi-overlap region. The semi-overlap region may vary depending on the size of the target and the direction of movement.

[0007] In the Chinese patent document numbered CN110766719B in the state of the art, a target tracking method and device is disclosed. In the Chinese patent document numbered CN115601392A in the state of the art, an obstacle motion detection method is disclosed.

[0008] In the European patent document numbered EP3682308A4 in the state of the art, a smart lidar system with low latency motion planning updates is disclosed.

[0009] This invention has been developed to prevent the plot point generated in the semioverlap region from producing speed distortion.

[0010] Objects and Brief Description of the Invention

[0011] It is an object of the present invention to provide a filter-controlled method in which velocity distortions in the semi-overlap region are avoided during target tracking.

[0012] It is a further object of the present invention to provide a method in which only the forward prediction step of the filter is used when the target is in the semi-overlap region.

[0013] It is a further object of the present invention to provide a method in which measurements are started to be generated after the target has completely left the overlap region, and the new measurements are associated with the forward prediction to ensure the continuity of the prediction without losing the target.

[0014] Detailed Description of the Invention

[0015] A system using the method for achieving the object of the present invention is shown in the accompanying figures.

[0016] These figures are;

[0017] Figure 1: A schematic representation of a system using the inventive method. The parts in the figure are numbered individually and the corresponding descriptions are given below.

[0018] 1. Sensor

[0019] 2. Target

[0020] 3. Obstacle

[0021] 4. Overlap zone

[0022] 5. Semi-overlap zone

[0023] 6. Field of view

[0024] 7. Overlap limit

[0025] 8. Sensor coverage area

[0026] A filter-controlled method for preventing velocity distortions in the overlap region according to the invention comprises the process steps,

[0027] Obtaining measurement data from a sensor (1),

[0028] Clustering of the incoming measurement data,

[0029] Generating plots as a result of clustering,

[0030] - Determining that the target (2) has entered the semi-overlap region (5) by comparing with the plots obtained from the previous step of the target tracking filter,

[0031] - Running the forward prediction and update phases of the tracking filter when the target is not in the semi-overlap region (5),

[0032] - Running only the forward prediction phase of the tracking filter when the target is in the overlap region (4).

[0033] In this study, a single plot is generated with multiple measurement data from the sensor (1) while the target (2) is in the field of view (6) and target kinematic prediction is performed with target tracking filters. Many different methods can be used for plot generation, and the aim of plot generation is to represent multiple measurements taken from a target (2) with a single point. For example, the center point of the target (2) detected by measurements can be used as a PLOT.

[0034] Sensor (1) can detect targets (2) within the sensor coverage area (8). If there is an obstacle (3) between the sensor (1) and the target (2), the target (2) will appear to be smaller in size due to fewer measurements taken by the sensor (1) when the target (2) starts to enter the semi-overlap region (5), resulting in a decrease in the predicted speed of the target (2). When the target (2) enters the overlap region (4) completely, the predicted velocity will be smaller than the actual velocity and the estimated position prediction of the target (2) will be behind the real one. An example of this velocity degradation is as follows;

[0035] Let the velocity of a target (2) of length 10m at time T be 2mA. For the entire length of the target (2) that can be detected with sensor (1), the plot point (the center point of the target) is 5m from the overlap boundary (7). (In this case, the end of the target (2) that is close to the overlap boundary (7) is at the full overlap boundary (7)) Since the 2m part of the target that enters the overlap region (4) after 1 second remains in the overlap region (4), sensor (1) will not be able to produce a measurement from this part of the target (2), while sensor (1) will produce a measurement for the 8m part of the target (2) that is not in the overlap region (4). The distance from the plot point produced for the measured part to the overlap zone (4) will be 4m. The velocity calculated using the measured plot will be approximately lower than the actual velocity, e.g. Im / s.

[0036] In order to solve the velocity distortion problem described in the example above, a solution is generated by using only the forward prediction step of the filter when the target (2) is in the semi-overlap region (5). With this solution, since no plot is generated as long as the target (2) is in the semi-overlap (5) and overlap region (6), there is an increase in the error area represented by the covariance matrix of the prediction by forward prediction. Although this may seem negative, since the velocity degradation is prevented, it is possible to correlate the current prediction with the plot generated when the target is completely out of the overlap region (4). Otherwise, the error area represented by the covariance matrix will not grow, but since the position obtained with the prediction will lag behind the target (2), the plot obtained when the target is completely out of the overlap region (4) may be outside the error area of the prediction and the new plot may be perceived as a different target (2).

[0037] The main objective of a target tracking filter is to predict the kinematic data (position, velocity, acceleration, etc.) of a target (2) from measurements taken from a sensor (1). The motion model of the target (2), the measurements taken from the sensor (1), the measurement errors of the generated plot, etc. determine the characteristics of the filter. Independently of these details, the kinematic parameters (position, velocity, acceleration, etc.) can be determined from several plots generated with measurements from the sensor (1). This phase is called the initialization phase, which will be done once. After initialization, each time a new plot is generated, the previously predicted target kinematics are advanced by the time elapsed according to the motion model and the new kinematics are calculated. This phase is called forward prediction. When a plot is generated, the existing forward kinematics data is updated using the plot data by Kalman filtering methods. This phase is also called the update phase.

[0038] When the target (2) is in the semi-overlap region (5), the forward estimation phase will be performed, and the update step with measurement will not be performed, since a plot cannot be generated within the scope of the invention.

Claims

CLAIMS1. A filter-controlled method for preventing velocity distortions in the overlap region characterised by comprising the process steps,Obtaining measurement data from a sensor (1), - Clustering of the incoming measurement data,Generating plots as a result of clustering,- Determining that the target (2) has entered the semi-overlap region (5) by comparing with the plots obtained from the previous step of the target tracking filter, - Running the forward prediction and update phases of the tracking filter when the target is not in the semi-overlap region (5),- Running only the forward prediction phase of the tracking filter when the target is in the overlap region (4).

Images