Object tracking device

Abstract

In this object tracking device, a predicted contour is calculated on the basis of a predetermined shape model of an object and estimated values calculated in the past, and a plurality of predicted values located on the predicted contour and / or within a predetermined range of the predicted contour are calculated independently of a plurality of observed values acquired by a sensor unit. And an association unit that associates each of the plurality of prediction values with at least one observation value of the plurality of observation values, generates a plurality of association sets, and calculates the current estimated value on the basis of the plurality of association sets generated by the association unit.

Description

[0001] CROSS-REFERENCE TO RELATED APPLICATIONS

[0002] This international application claims priority based on Japanese Patent Application No. 2023-024583 filed with the Japan Patent Office on February 20, 2023, the entire contents of which are incorporated herein by reference. Technical Field

[0003] The present disclosure relates to a technology for a mobile object to track surrounding objects. Background Art

[0004] In Extended Object Tracking (EOT), described in Non-Patent Document 1, the radar device assumes that the reflected signal is generated by the outline of a target object. Predicted values ​​on the predicted outline of the target object are associated with observed values, and the target object's motion is estimated in a time series. Specifically, in EOT, the intersection points of multiple straight lines with the predicted outline are calculated as predicted values. These straight lines pass through the center of the predicted outline and each of the multiple observed values. Furthermore, each of these calculated predicted values ​​is associated with the observed value on the same straight line as the predicted value.

[0005] Non-Patent Literature 1: Tokizawa Soichiro, Yeda Yoshisuke, and Suganuma Naoki, "Extended Object Tracking with Improved Stability and Real-Time Performance for Autonomous Driving," Proceedings of the Automotive Technology Conference, Vol. 52, No. 5, September 2021. Summary of the Invention

[0006] The inventors' detailed research revealed that, in the aforementioned EOT, predicted values ​​are calculated based on observed values. Therefore, if the predicted values ​​are associated with erroneous observed values, the estimation of the target object's motion state is continuously affected by this erroneous association. Furthermore, the inventors discovered that the accuracy of the estimated target object position is reduced.

[0007] Preferably, a technical solution of the present disclosure can provide an object tracking device capable of suppressing a decrease in the accuracy of estimating the position of an object.

[0008] An object tracking device according to one technical solution of the present disclosure includes a sensor unit, a contour calculation unit, a predicted value calculation unit, an association unit, and an inference unit. The sensor unit is mounted on a mobile object. The sensor unit transmits and receives sensor waves around the mobile object, acquiring multiple observation values. The multiple observation values ​​correspond to different reflection positions. The contour calculation unit calculates a predicted contour based on a predetermined object shape model and previously calculated estimated values. The estimated value is an estimated value representing the state of the target object, including the position and orientation of the object. The predicted contour is a predicted value representing the current contour of the target object. The predicted value calculation unit calculates multiple predicted values ​​within a specified range located on and / or inside the predicted contour, independently of the multiple observation values ​​acquired by the sensor unit. The association unit associates each of the multiple predicted values ​​with at least one of the multiple observation values ​​to generate multiple association sets. The inference unit calculates the current estimated value based on the multiple association sets generated by the association unit.

[0009] An object tracking device according to one aspect of the present disclosure calculates multiple predicted values ​​independently of multiple observed values, thereby preventing the erroneous association between predicted values ​​and observed values ​​from persisting and suppressing a decrease in the accuracy of object position estimation. BRIEF DESCRIPTION OF THE DRAWINGS

[0010] Figure 1 This is a block diagram showing the hardware configuration of the object tracking device according to the first embodiment.

[0011] Figure 2 This is a block diagram showing the functional configuration of the object tracking device according to the first embodiment.

[0012] Figure 3 This is a diagram showing an example of the mounting position and detection range of the sensor unit according to the first embodiment in the vehicle.

[0013] Figure 4 This is a diagram showing another example of the mounting position and detection range of the sensor unit according to the first embodiment in the vehicle.

[0014] Figure 5 This is a flowchart showing the tracking process executed by the object tracking device according to the first embodiment.

[0015] Figure 6 This is a flowchart showing the predicted value calculation process executed by the object tracking device according to the first embodiment.

[0016] Figure 7 This is a flowchart showing the estimation process executed by the object tracking device according to the first embodiment.

[0017] Figure 8It is a diagram showing the prediction profile, observation values, and prediction values ​​according to the first embodiment.

[0018] Figure 9 This is a diagram showing the association between one predicted value and a plurality of observed values ​​according to the first embodiment.

[0019] Figure 10 This is a diagram showing a state in which a predicted value is calculated in a direct reflection area on the outline of an ellipse according to the first embodiment.

[0020] Figure 11 This is a diagram showing a state in which a predicted value is calculated for a direct reflection area on a rectangular outline according to the first embodiment.

[0021] Figure 12 This is a diagram showing a state in which a predicted value is calculated in a direct reflection area on an elliptical outline when a sensor wave is irradiated from behind an object in the first embodiment.

[0022] Figure 13 This is a diagram showing a state in which predicted values ​​are calculated in a direct reflection area on an elliptical outline when a sensor wave is irradiated from the side of an object in the first embodiment.

[0023] Figure 14 This is a diagram showing a state in which the positional intervals of the predicted values ​​become narrower as the distance approaches the center of the predetermined range on the outline of the ellipse according to the first embodiment.

[0024] Figure 15 This is a diagram showing a state in which predicted values ​​are calculated on and within the outline of an ellipse when a sensor wave is irradiated from behind an object in the first embodiment.

[0025] Figure 16 This is a flowchart showing the predicted value calculation process executed by the object tracking device according to the second embodiment.

[0026] Figure 17 This is a flowchart showing the estimation process executed by the object tracking device according to the second embodiment.

[0027] Figure 18 This is a flowchart showing a predicted value calculation process executed by the object tracking device according to the third embodiment. DETAILED DESCRIPTION

[0028] (First embodiment)

[0029] 1. Structure

[0030] Reference Figure 12. The structure of the object tracking device 10 according to this embodiment will be described. Object tracking device 10 includes a sensor unit 11 and a processing device 20, and is mounted on a vehicle 50. Sensor unit 11 is a radar, a lidar, a sonar, or the like. Radar transmits radio waves such as millimeter waves as sensor waves and receives reflected waves generated by the radar waves reflecting off objects. Lidar transmits light as sensor waves and receives reflected waves generated by the light reflecting off objects. Sonar transmits sound waves as sensor waves and receives reflected waves generated by the sound waves reflecting off objects.

[0031] like Figure 3 As shown, the sensor unit 11 may be mounted in the front center of the vehicle 50 (for example, the center of the front bumper) and have a detection area A1 in the front center of the vehicle 50. Alternatively, Figure 4 As shown, the sensor unit 11 is mounted not only at the front center of the vehicle 50, but also at the front left, front right, rear left, and rear right of the vehicle 50 (for example, at the left and right ends of the front bumper and the left and right ends of the rear bumper). In other words, the sensor unit 11 may have detection areas A2 at the front left, front right, rear left, and rear right of the vehicle 50 in addition to the detection area A1. The sensor unit 11 only needs to be mounted at least one of the front center, front left, front right, rear left, and rear right of the vehicle 50.

[0032] The processing device 20 includes a microcomputer including a CPU, ROM, RAM, etc. The processing device 20 realizes the functions of the contour calculation unit 13, the predicted value calculation unit 15, the association unit 17, and the inference unit 19 by executing a program stored in a non-migrating physical recording medium through the CPU. By having the various functions described above, the processing device 20 performs tracking processing to infer the motion state of an object according to a time series. The details of the various functions described above will be described later. In this embodiment, the ROM is equivalent to the non-migrating physical recording medium. In addition, some or all of the various functions implemented by the processing device 20 can also be implemented using hardware that combines logic circuits, analog circuits, etc.

[0033] <2. Processing>

[0034] <2-1. Tracking Process>

[0035] Reference Figure 5 The tracking process performed by object tracking device 10 will be described with reference to the flowchart of FIG. Object tracking device 10 repeatedly performs this tracking process in a predetermined processing cycle.

[0036] In S10, as Figure 9As shown, the sensor unit 11 transmits sensor waves to the periphery of the vehicle 50 and receives reflected waves generated by reflection from an object. Furthermore, the sensor unit 11 obtains a plurality of observation values ​​P3 based on the received reflected waves. The sensor unit 11 is a high-resolution sensor. By transmitting the sensor wave once, it is possible to obtain a plurality of reflected waves reflected at different reflection positions (i.e., reflection points) of a single object and obtain an observation value P3 based on each reflected wave. Therefore, the plurality of observation values ​​P3 respectively correspond to different reflection positions of a single object. The plurality of observation values ​​P3 respectively include a reflection position (specifically, the distance from the sensor unit 11 to the reflection point, and the orientation of the reflection point relative to the sensor unit 11) as a physical quantity. In addition to including the reflection position, the plurality of observation values ​​P3 may also include a relative speed as a physical quantity.

[0037] Next, in S20, the contour calculation unit 13 performs a prediction process to calculate a predicted contour L1 of the target object. The predicted contour L1 is a predicted value of the contour of the modeled target object shape.

[0038] As described above, the sensor unit 11 obtains multiple observation values ​​P3 from a single object. The shape of the object can be inferred based on the distribution of the multiple observation values ​​P3. Therefore, the object tracking device 10 can generate an object that has a shape in addition to a motion state based on the multiple observation values ​​P3. The shape here is different from a point mass that does not have an area, and has an area. The object tracking device 10 uses the shape model of the object to perform extended object tracking (hereinafter, EOT). Extended object tracking is a method of modeling the object assuming that the object has a shape and inferring the motion state of the object in a time series.

[0039] like Figure 9 As shown, the contour calculation unit 13 calculates the predicted contour L1 based on the predetermined shape model of the object and the estimated value P2 calculated in the past processing cycle (specifically, the last processing cycle). In this embodiment, the object tracked by the object tracking device 10 is a car (specifically, a four-wheeled vehicle). Therefore, as the shape model of the object, the contour calculation unit 13 uses a circle model, Figure 10 The elliptical model shown, Figure 11 Alternatively, the outline calculation unit 13 may estimate the size and shape of the target object based on the distribution of the plurality of observation values ​​P3 and select and use an appropriate model from a plurality of pre-prepared models.

[0040] The estimated value P2 is a value obtained by inferring the state of the target object. The state of the target object includes its position and orientation. The estimated value P2 includes at least one physical quantity at the target object's reference point. The object tracking device 10 infers the motion state of the reference point in a time series based on multiple observation values ​​P3, multiple prediction values ​​P4, a shape model, and a predetermined filter, and calculates the estimated value P2. The predetermined filter is, for example, a Kalman filter. The estimated value P2 includes, for example, the x-direction position and y-direction position of the reference point, its velocity, its direction of travel, and its angular velocity. The x-direction corresponds to the length of the vehicle 50, and the y-direction corresponds to the width of the vehicle 50.

[0041] In this embodiment, the reference point is the axle center of the vehicle's rear wheels. The contour calculation unit 13 calculates a predicted contour L1 of the target object's shape for the current processing cycle based on the past estimated values ​​P2 and the shape model. The interior of predicted contour L1 corresponds to the predicted target object's location area for the current processing cycle. In other words, the contour calculation unit 13 predicts the target object's location area for the current processing cycle based on the past estimated values ​​P2 and the shape model. Therefore, there is a high probability that the predicted contour L1 will be calculated near the multiple observation values ​​P3 acquired in S10.

[0042] Next, in S30, the predicted value calculation unit 15 performs a predicted value calculation process to calculate a plurality of predicted values ​​P4 located on and / or within the predicted profile L1 calculated in S20, independently of the plurality of observed values ​​P3. That is, the predicted value calculation unit 15 calculates the plurality of predicted values ​​P4 on and / or within the predicted profile L1 using a calculation method that does not depend on the plurality of observed values ​​P3 (i.e., independently of the plurality of observed values ​​P3).

[0043] exist Figure 8 In the reference example shown, multiple predicted values ​​P4 are calculated based on multiple observed values ​​P3. Specifically, in this reference example, it is assumed that sensor waves are reflected from the outline of the target object. Straight lines are drawn through the center point P1 and each of the multiple observed values ​​P3. The intersection of each straight line with the predicted contour L1 is calculated as a predicted value P4, and the predicted values ​​P4 on the same straight line are associated with the observed values ​​P3. The center point P1 is the center of the predicted contour L1.

[0044] like Figure 8 As shown in FIG, when the predicted contour L1 deviates from the actual contour LL of the target object, the predicted value P4 is associated with the observed value P3 that does not correspond to the predicted value P4. Figure 8In the example, the predicted value P4 for the left side of the vehicle is associated with the observed value P3 for the right side of the vehicle. If multiple predicted values ​​P4 are calculated based on multiple observed values ​​P3, a single incorrect association could persist. This could reduce the accuracy of the calculated estimated value P2, leading to a decrease in object tracking accuracy.

[0045] Therefore, in this embodiment, the predicted value calculation unit 15 calculates a plurality of predicted values ​​P4 within a predetermined range on and / or within the predicted contour L1 independently of the plurality of observed values ​​P3. The details of the predicted value calculation process will be described later.

[0046] Next, in S40, the associating unit 17 associates the multiple predicted values ​​P4 calculated in S30 with one or more of the multiple observed values ​​P3 obtained in S10. In the present embodiment, the associating unit 17 associates the multiple predicted values ​​P4 with two or more of the multiple observed values ​​P3 to generate multiple association sets. In detail, the associating unit 17 associates one predicted value P4 with two or more observed values ​​P3 within a set range of the predicted value P4 to generate one association set. Therefore, each association set includes one predicted value P4 and two or more observed values ​​P3 associated with the predicted value P4. In this way, by allowing two or more associations to be established for one predicted value P4, the possibility of including a correct association in the two or more associations becomes higher. Therefore, by allowing multiple associations to be established, the inferred value P2 is affected by the correct association during object tracking, and the inferred value P2 converges to a correct value. When the associating unit 17 associates one predicted value P4 with one observed value P3, if the association is erroneous, there is a possibility that the erroneous association will continue.

[0047] Next, in S50, the estimation unit 19 performs an estimation process, applies a filter such as a Kalman filter to the plurality of associated sets calculated in S40, and calculates the current estimated value P2. The details of the estimation process will be described later.

[0048] <2-2. Prediction value calculation process>

[0049] Next, refer to Figure 6 The flowchart of will now be used to explain the details of the predicted value calculation process performed by the predicted value calculation unit 15.

[0050] In S100, the predicted value calculation unit 15 determines whether the distance to the target object is greater than or equal to a threshold. This distance is the distance of the estimated value P2 calculated in the previous processing cycle, or the distance calculated based on the position of the estimated value P2. Alternatively, this distance may be the distance calculated based on the plurality of observed values ​​P3 acquired in S10.

[0051] When the target object is located far from the sensor unit 11, the positions of the object's reflection points vary significantly, with the reflection points being widely distributed throughout the entire area of ​​the object. Therefore, when the target object is located far from the sensor unit 11 and the predicted value calculator 15 calculates multiple predicted values ​​P4 within a predetermined range on and / or within the predicted contour L1, narrowing the predetermined range may reduce the accuracy of associating a single predicted value P4 with two or more observed values ​​P3. Consequently, the calculation accuracy of the estimated value P2 may be reduced.

[0052] On the other hand, when a target object is located near the sensor unit 11, the positional variation of the object's reflection points is small, and the reflection points are concentrated in a specific area of ​​the object. Therefore, if the target object is located near the sensor unit 11 and the predicted value calculation unit 15 calculates multiple predicted values ​​P4 within a predetermined range on and / or within the predicted contour L1, expanding the predetermined range may reduce the accuracy of associating a single predicted value P4 with two or more observed values ​​P3. Consequently, the calculation accuracy of the estimated value P2 may be reduced.

[0053] Therefore, the predicted value calculation unit 15 changes the predetermined range according to the distance of the target object. If the predicted value calculation unit 15 determines in S100 that the distance is greater than the threshold, it proceeds to S110, and if it determines that the distance is less than the threshold, it proceeds to S120.

[0054] In S110, if Figure 9 As shown, the predicted value calculation unit 15 sets the prescribed range around the entire predicted contour L1. Furthermore, the predicted value calculation unit 15 calculates a plurality of predicted values ​​P4 at the determined position intervals within the prescribed range. The plurality of predicted values ​​P4 each have a coordinate value on the predicted contour L1. The determined position intervals may also be equal intervals. Alternatively, the determined position intervals may be smaller the closer to the center of the prescribed range, so that more predicted values ​​P4 closer to the center of the prescribed range contribute to the update of the inferred value P2 (see Figure 14 ).

[0055] In S120, the predicted value calculation unit 15 calculates a direct reflection area in the predicted profile L1. The direct reflection area corresponds to an area where the sensor unit 11 can directly irradiate the sensor wave.

[0056] like Figure 10 As shown, when the shape model is an elliptical model or a circular model, the predicted value calculation unit 15 calculates the area on the side close to the sensor unit 11 between the first point Pa and the second point Pb as the direct reflection area. The first point Pa is a point on the predicted contour L1 and is the point of intersection of the first tangent line La passing through the sensor unit 11 and the shape model. The second point Pb is a point on the predicted contour L1 different from the first point Pa and is the point of intersection of the second tangent line Lb passing through the sensor unit 11 and the shape model. Figure 12 The direct reflection area when the sensor unit 11 irradiates the sensor wave toward the rear of an object (specifically, a vehicle) is shown. Figure 13 The direct reflection area when the sensor unit 11 irradiates the sensor wave to the side of the object is shown.

[0057] In addition, if Figure 11 As shown, when the shape model is a rectangular model including a first vertex PP1, a second vertex PP2, a third vertex PP3, a fourth vertex PP4, a first side SS1, and a second side SS2, the predicted value calculation unit 15 calculates the first side SS1 and the second side SS2 as the direct area. The first vertex PP1 is the farthest vertex from the sensor unit 11 among the four vertices of the rectangular model. The second vertex PP2 and the third vertex PP3 are two vertices adjacent to the first vertex PP1. The fourth vertex PP4 is between the second vertex PP2 and the third vertex PP3 and is a different vertex from the first vertex PP1. The first side SS1 connects the fourth vertex PP4 to the second vertex PP2. The second side SS2 connects the fourth vertex PP4 to the third vertex PP3.

[0058] In S130, the prediction value calculation unit 15 sets the direct reflection area as a predetermined range, and calculates a plurality of prediction values ​​P4 at predetermined position intervals within the predetermined range. Figure 10 As shown in FIG11, the determined position intervals may also be equal intervals. Alternatively, Figure 14 As shown, the determined position intervals are smaller as they are closer to the center of the predetermined range so that the contribution of the predicted value P4 to the estimated value P2 increases as it is closer to the center of the predetermined range.

[0059] <2-3. Inference Process>

[0060] Next, refer to Figure 7 The flowchart of will be used to explain the details of the inference processing performed by the inference unit 19.

[0061] In S200, the inference unit 19 determines whether the subsequent processes of S210 to S230 have been executed for all predicted values ​​P4 calculated in S20. If the inference unit 19 determines that the processes of S210 to S230 have not been executed for all predicted values ​​P4, it selects a predicted value P4 from the predicted values ​​P4 determined not to have been executed for S210 to S230 and proceeds to S210. If the inference unit 19 determines that the processes of S210 to S230 have been executed for all predicted values ​​P4, it proceeds to S240.

[0062] In S210, the inference unit 19 determines whether the predicted value P4 selected in S200 is within the direct reflection area. If the specified range is set to the entire perimeter of the predicted profile L1, the specified range includes both the direct reflection area and the indirect reflection area. The indirect reflection area is an area where the sensor unit 11 cannot directly illuminate the sensor wave. If the inference unit 19 determines that the selected predicted value P4 is within the direct reflection area, the process proceeds to S220. If the inference unit 19 determines that the selected predicted value P4 is within the indirect reflection area, the process skips S220 and proceeds to S230.

[0063] In S220, the inference unit 19 increases the contribution of the selected predicted value P4 to the update of the estimated value P2 by a greater degree than a reference value. For example, the inference unit 19 increases the contribution by adding a predetermined value to the reference value of the contribution. When the selected predicted value P4 is located in a non-direct reflection area, the contribution is the reference value. The associated set that includes the predicted value P4 located in a direct reflection area is more reliable than the associated set that includes the predicted value P4 located in a non-direct reflection area. Therefore, when the selected predicted value P4 is located in a direct reflection area, the inference unit 19 increases the contribution to the update of the estimated value P2.

[0064] In S230, the estimation unit 19 calculates the update amount of the estimated value P2 by the observation value associated with the selected predicted value P4. Specifically, the estimation unit 19 calculates the update amount using an extended Kalman filter, which is a nonlinear filter.

[0065] In S240 , the update amount of the estimated value P2 calculated for each predicted value P is weighted averaged and added according to the contribution, thereby updating the estimated value P2. The updated estimated value P2 becomes the estimated value P2 in the current processing cycle.

[0066] <3. Effect>

[0067] According to the first embodiment described in detail above, the following effects are achieved.

[0068] (1) In object tracking device 10, multiple predicted values ​​P4 are calculated independently of multiple observed values ​​P3. This prevents the predicted value P4 from being erroneously associated with at least one observed value P3, thereby suppressing a decrease in the accuracy of object position estimation.

[0069] (2) One predicted value P4 is associated with multiple observed values ​​P3. By allowing multiple associations to be established for one predicted value P4, the correct association can be included among the multiple associations, allowing the estimated position of the target object to converge to the correct observed position. This, in turn, improves the accuracy of the estimated target object position.

[0070] (3) When a plurality of predicted values ​​P4 are calculated at predetermined position intervals within a predetermined range, the calculation process of the plurality of predicted values ​​P4 can be simplified.

[0071] (4) When a plurality of predicted values ​​P4 are calculated at equal positional intervals within a predetermined range, the calculation process of the plurality of predicted values ​​P4 can be simplified.

[0072] (5) The closer to the center of the predetermined range, the higher the probability of obtaining more observation values ​​P3. Therefore, by calculating the predicted values ​​P4 more densely as the position approaches the center of the predetermined range, the estimation accuracy of the target object position can be improved.

[0073] (6) Many observation values ​​P3 are obtained from the direct reflection area. Therefore, by setting the predetermined range as the direct reflection area, the estimation accuracy of the target object position can be improved.

[0074] (7) When a circle model or an ellipse model is used as the shape model, the direct reflection area can be simply calculated by calculating the area between the first point Pa and the second point Pb as the direct reflection area.

[0075] (8) When a rectangular model is used as a shape model, the direct reflection area can be simply calculated by calculating the first side SS1 and the second side SS2 as the direct reflection area.

[0076] (9) When the predetermined range is set to the entire surrounding area, the contribution of the predicted value P4 in the direct reflection area to the update of the estimated value P2 is calculated to be greater than the contribution of the predicted value P4 in the indirect reflection area to the update of the estimated value P2. This improves the accuracy of the estimated position of the target object.

[0077] (10) The predetermined range is set to the entire periphery of the predicted contour L1 in the distance from the sensor unit 11, and the predetermined range is set to the direct reflection area in the vicinity of the sensor unit 11. This allows the position of the target object to be estimated with high accuracy from the vicinity of the sensor unit 11 to the distance.

[0078] (Second embodiment)

[0079] <1. Differences from the First Embodiment>

[0080] The basic structure of the second embodiment is the same as that of the first embodiment, so the following describes the differences. In addition, the same reference numerals as those in the first embodiment represent the same structures, and reference is made to the previous description.

[0081] In the first embodiment described above, the prescribed range is changed according to the distance to the target object. In contrast, the second embodiment differs from the first embodiment in that the prescribed range is set regardless of the distance to the target object. Specifically, in the second embodiment, the predicted value calculation unit 15 sets the direct reflection area to the prescribed range regardless of the distance to the target object. The inference unit 19 calculates the estimated value P4 without considering the contribution of the predicted value P4 to the update of the estimated value P2. Therefore, in the second embodiment, the predicted value calculation unit 15 performs Figure 16 The flowchart is replaced by Figure 6 In the second embodiment, the inference unit 19 executes Figure 17 The flowchart is replaced by Figure 7 Flowchart of the process.

[0082] <2. Processing>

[0083] <2-1. Prediction value calculation process>

[0084] Reference Figure 16 The flowchart of FIG. 1 illustrates the predicted value calculation process executed by the predicted value calculation unit 15 .

[0085] In S310 , the predicted value calculation unit 15 calculates the direct reflection area in the predicted outline L1 in the same manner as in S120 .

[0086] Next, in S320, similar to S130, the predicted value calculation unit 15 sets the direct reflection area to a predetermined range and calculates a plurality of predicted values ​​P4 at predetermined positional intervals within the predetermined range. That is, in this embodiment, the predicted value calculation unit 15 calculates a plurality of predicted values ​​P4 within the direct reflection area regardless of the distance to the target object.

[0087] <2-2. Inference Process>

[0088] Next, refer to Figure 17 The flowchart of FIG. 1 illustrates the inference processing performed by the inference unit 19.

[0089] In S500, similar to S200, the inference unit 19 determines whether the subsequent process of S510 has been executed for all predicted values ​​P4 calculated in S20. If the inference unit 19 determines that the process of S510 has not been executed for all predicted values ​​P4, it selects one predicted value P4 from the predicted values ​​P4 determined not to have been executed for S510 and proceeds to the process of S510. If the inference unit 19 determines that the process of S510 has been executed for all predicted values ​​P4, it proceeds to the process of S520.

[0090] In S510 , the estimation unit 19 calculates the update amount of the estimated value P2 by the observation value associated with the selected predicted value P4 , similarly to S230 .

[0091] In S520, the estimation unit 19 averages the update amount of the estimated value P4 calculated for each predicted value P4 and updates the estimated value P2. That is, in this embodiment, the estimation unit 19 updates the estimated value P2 by giving equal weights to all predicted values ​​P4.

[0092] According to the second embodiment described above, the same effects as the above-mentioned effects (1) to (8) are achieved.

[0093] (Third embodiment)

[0094] <1. Differences from the Second Embodiment>

[0095] The basic structure of the third embodiment is the same as that of the second embodiment, so the following describes the differences. In addition, the same reference numerals as those of the second embodiment represent the same structures, and reference is made to the previous description.

[0096] In the second embodiment described above, the direct reflection area is set to a predetermined range regardless of the distance to the target object. In contrast, the third embodiment differs from the second embodiment in that the entire periphery of the predicted contour L1 is fixed to a predetermined range regardless of the distance to the target object. The estimation unit 19 calculates the estimated value P4 without considering the contribution of the predicted value P4 to the update of the estimated value P2. Therefore, in the third embodiment, the predicted value calculation unit 15 executes Figure 18 The flowchart is replaced by Figure 16 In the third embodiment, similarly to the second embodiment, the inference unit 19 executes Figure 17 Flowchart of the process.

[0097] <2. Prediction value calculation processing>

[0098] Reference Figure 18 The flowchart of will now be used to explain the predicted value calculation process performed by the predicted value calculation unit 15.

[0099] In S400, the predicted value calculation unit 15 sets the predetermined range to the entire perimeter of the predicted contour L1 and calculates a plurality of predicted values ​​P4 at predetermined positional intervals within the predetermined range. Specifically, in this embodiment, the predicted value calculation unit 15 calculates a plurality of predicted values ​​P4 for the entire perimeter of the predicted contour L1, regardless of the distance to the target object.

[0100] According to the third embodiment described above, the same effects as the above-mentioned effects (1) to (5) and (9) are achieved, and the following effects (11) and (12) are achieved.

[0101] (11) By fixing the predetermined range, the calculation process of the plurality of predicted values ​​P4 can be simplified.

[0102] (12) By setting the predetermined range to the entire periphery of the predicted contour L1, it is possible to suppress the estimated position of the target object from deviating significantly from the true position.

[0103] (Other Embodiments)

[0104] As mentioned above, although embodiment of this disclosure was described, this disclosure is not limited to the said embodiment, It can be implemented with various deformation|transformation.

[0105] (a) When the predetermined specified range is set to the entire perimeter of predicted profile L1, the contribution of predicted value P4 in the direct reflection area to the update of estimated value P2 can be made greater than the contribution of predicted value P4 in the indirect reflection area to the update of estimated value P2. Alternatively, the predetermined specified range can be set to the range corresponding to one or more tires on predicted profile L1. Since tires tend to have more reflection points, setting the range corresponding to the tires as the specified range allows for high-precision estimation of the target object's position.

[0106] (b) In the above embodiment, the predicted value calculation unit 15 changes the prescribed range according to the distance of the target object, but the prescribed range may be changed according to the state of the target object. In the case where the sensor unit 11 irradiates the sensor wave to the side of the vehicle being tracked, almost no observation value P3 corresponding to the reflection position inside the vehicle is obtained. On the other hand, in the case where the sensor unit 11 irradiates the sensor wave to the rear of the vehicle being tracked, relatively more observation values ​​P3 corresponding to the reflection position inside the vehicle are obtained. Therefore, the predicted value calculation unit 15 may also change the prescribed range according to the direction of the vehicle being tracked. For example, Figure 15As shown, when the sensor unit 11 emits sensor waves behind the vehicle being tracked, the predicted value calculation unit 15 sets the predetermined range on and within the predicted contour L1. Alternatively, when the sensor unit 11 emits sensor waves to the side of the vehicle being tracked, the predicted value calculation unit 15 may set the predetermined range only on the predicted contour L1. By dynamically setting the predetermined range based on the state of the target object, the accuracy of the target object position estimation can be improved.

[0107] (c) In the above embodiment, the object tracking device 10 is mounted on an automobile, but it may be mounted on a mobile object other than an automobile. For example, the object tracking device 10 may be mounted on a train, ship, aircraft, motorcycle, drone, or other mobile object.

[0108] (d) The object tracking device 10 and the method thereof described in the present disclosure may also be implemented by a dedicated computer provided by a processor and a memory programmed to execute one or more functions embodied by a computer program. Alternatively, the object tracking device 10 and the method thereof described in the present disclosure may also be implemented by a dedicated computer provided by a processor composed of one or more dedicated hardware logic circuits. Alternatively, the object tracking device 10 and the method thereof described in the present disclosure may also be implemented by one or more dedicated computers composed of a combination of a processor and a memory programmed to execute one or more functions and a processor composed of one or more hardware logic circuits. In addition, a computer program may also be stored as an instruction executed by a computer in a non-portable tangible recording medium that can be read by a computer. The method for implementing the functions of each part included in the object tracking device 10 does not necessarily include software, and all of its functions may also be implemented using one or more hardware.

[0109] (e) It is also possible to implement multiple functions of a component of the above embodiment by multiple components, or to implement a function of a component by multiple components. In addition, it is also possible to implement multiple functions of multiple components by one component, or to implement a function implemented by multiple components by one component. In addition, part of the structure of the above embodiment may be omitted. In addition, at least part of the structure of the above embodiment may be added to or replaced with the structure of another above embodiment.

[0110] (f) In addition to the object tracking device 10 described above, the present disclosure may be implemented in various ways, such as a system having the object tracking device 10 as a component, a program for causing a computer to function as the object tracking device, a non-migrating physical recording medium such as a semiconductor memory having the program recorded thereon, and an object tracking method.

[0111] [Technical Ideas Disclosed in This Specification]

[0112] [Project 1]

[0113] An object tracking device comprising:

[0114] A sensor unit (11) is mounted on a moving body (50) and is configured to transmit and receive sensor waves to and around the moving body, and obtain a plurality of observation values, wherein the plurality of observation values ​​correspond to different reflection positions;

[0115] A contour calculation unit (13) is configured to calculate a predicted contour based on a predetermined object shape model and an estimated value calculated in the past, wherein the estimated value is an estimated value indicating a state of an object including a position and an orientation of the object, and the predicted contour is a predicted value indicating a current contour of the object;

[0116] A predicted value calculation unit (15) is configured to calculate a plurality of predicted values ​​within a predetermined range located on and / or inside the predicted contour, independently of the plurality of observation values ​​acquired by the sensor unit;

[0117] an associating unit (17) configured to associate each of the plurality of predicted values ​​with at least one of the plurality of observed values ​​to generate a plurality of associated sets; and

[0118] The inference unit (19) is configured to calculate the current inference value based on the plurality of association sets generated by the association unit.

[0119] [Project 2]

[0120] The object tracking device according to item 1, wherein:

[0121] The associating unit is configured to associate each of the plurality of predicted values ​​with two or more observed values ​​from the plurality of observed values.

[0122] The plurality of association sets respectively include one predicted value among the plurality of predicted values ​​and two or more observed values ​​among the plurality of observed values.

[0123] [Item 3]

[0124] The object tracking device according to item 1 or 2, wherein:

[0125] The above-mentioned prescribed range is predetermined.

[0126] [Item 4]

[0127] The object tracking device according to item 1 or 2, wherein:

[0128] The predicted value calculation unit is configured to set the predetermined range according to a state of the target object.

[0129] [Item 5]

[0130] The object tracking device according to any one of items 1 to 4, wherein:

[0131] The predicted value calculation unit is configured to calculate the plurality of predicted values ​​at predetermined positional intervals within the predetermined range.

[0132] [Item 6]

[0133] The object tracking device according to any one of items 1 to 5, wherein:

[0134] The predicted value calculation unit is configured to calculate the plurality of predicted values ​​at equal positional intervals within the predetermined range.

[0135] [Item 7]

[0136] The object tracking device according to any one of items 1 to 6, wherein:

[0137] The predicted value calculation unit is configured to calculate the plurality of predicted values ​​such that positional intervals decrease as the predicted values ​​approach the center of the predetermined range.

[0138] [Item 8]

[0139] The object tracking device according to any one of items 1 to 7, wherein:

[0140] The above-mentioned prescribed range is the entire periphery of the above-mentioned predicted contour.

[0141] [Item 9]

[0142] The object tracking device according to any one of items 1, 2, 4 to 7, wherein:

[0143] The predetermined range is a range in which the sensor portion can be directly irradiated with the sensor wave.

[0144] [Item 10]

[0145] The object tracking device according to item 9, wherein:

[0146] The above shape model is a circle model or an ellipse model.

[0147] The predetermined range is a region between the first point and the second point on the predicted contour, which is closer to the sensor portion.

[0148] The first point is a junction point of a first tangent line and the predicted contour, the first tangent line passes through the sensor portion and is tangent to the predicted contour,

[0149] The second point is a point of contact between a second tangent line and the predicted contour and is a point different from the first point. The second tangent line passes through the sensor portion and is tangent to the predicted contour.

[0150] [Item 11]

[0151] The object tracking device according to item 9, wherein:

[0152] The shape model is a rectangular model having a first vertex, a second vertex, a third vertex, a fourth vertex, a first side, and a second side, wherein the first vertex is farthest from the sensor portion, the second vertex and the third vertex are adjacent to the first vertex, the fourth vertex is located between the second vertex and the third vertex, the first side connects the fourth vertex to the second vertex, and the second side connects the fourth vertex to the third vertex.

[0153] The prescribed range includes the first side and the second side.

[0154] [Item 12]

[0155] The object tracking device according to item 8, wherein:

[0156] The predetermined range includes a first range in which the sensor unit can directly irradiate the sensor wave, and a second range in which the sensor unit cannot directly irradiate the sensor wave.

[0157] The plurality of predicted values ​​include a first predicted value within the first range and a second predicted value within the second range.

[0158] The estimating unit is configured to make the first contribution higher than the second contribution.

[0159] The first contribution is a contribution of the first predicted value to the update of the inferred value, and the second contribution is a contribution of the second predicted value to the update of the inferred value.

[0160] [Item 13]

[0161] The object tracking device according to any one of items 1 to 12, wherein:

[0162] The predicted value calculation unit is configured to change the predetermined range according to the distance of the target object.

[0163] [Item 14]

[0164] The object tracking device according to item 13, wherein:

[0165] The above-mentioned prediction value calculation unit is composed of:

[0166] When the distance to the target object is greater than or equal to a predetermined distance, the entire periphery of the predicted contour is set as the predetermined range.

[0167] When the distance to the target object is less than a predetermined distance, the range in which the sensor unit can directly irradiate the sensor wave is set to the predetermined range.

Claims

1. An object tracking device comprising: A sensor unit (11) is mounted on a moving body (50) and is configured to transmit and receive sensor waves to and around the moving body, and obtain a plurality of observation values, wherein the plurality of observation values ​​correspond to different reflection positions; A contour calculation unit (13) is configured to calculate a predicted contour based on a predetermined object shape model and an estimated value calculated in the past, wherein the estimated value is an estimated value indicating a state of an object including a position and an orientation of the object, and the predicted contour is a predicted value indicating a current contour of the object; A predicted value calculation unit (15) is configured to calculate a plurality of predicted values ​​within a predetermined range located on and / or inside the predicted contour, independently of the plurality of observation values ​​acquired by the sensor unit; an associating unit (17) configured to associate each of the plurality of predicted values ​​with at least one of the plurality of observed values ​​to generate a plurality of associated sets; and The inference unit (19) is configured to calculate the current inference value based on the plurality of association sets generated by the association unit.

2. The object tracking device according to claim 1, wherein: The associating unit is configured to associate each of the plurality of predicted values ​​with two or more observed values ​​from the plurality of observed values. The plurality of association sets respectively include one predicted value among the plurality of predicted values ​​and two or more observed values ​​among the plurality of observed values.

3. The object tracking device according to claim 1 or 2, wherein: The above-mentioned prescribed range is predetermined.

4. The object tracking device according to claim 1 or 2, wherein: The predicted value calculation unit is configured to set the predetermined range according to a state of the target object.

5. The object tracking device according to claim 1 or 2, wherein: The predicted value calculation unit is configured to calculate the plurality of predicted values ​​at predetermined positional intervals within the predetermined range.

6. The object tracking device according to claim 1 or 2, wherein: The predicted value calculation unit is configured to calculate the plurality of predicted values ​​at equal positional intervals within the predetermined range.

7. The object tracking device according to claim 1 or 2, wherein: The predicted value calculation unit is configured to calculate the plurality of predicted values ​​such that positional intervals decrease as the predicted values ​​approach the center of the predetermined range.

8. The object tracking device according to claim 5, wherein: The above-mentioned prescribed range is the entire periphery of the above-mentioned predicted contour.

9. The object tracking device according to claim 5, wherein: The predetermined range is a range in which the sensor portion can be directly irradiated with the sensor wave.

10. The object tracking device according to claim 9, wherein: The above shape model is a circle model or an ellipse model. The predetermined range is a region between the first point and the second point on the predicted contour, which is closer to the sensor portion. The first point is a junction point of a first tangent line and the predicted contour, the first tangent line passes through the sensor portion and is tangent to the predicted contour, The second point is a point of contact between a second tangent line and the predicted contour and is a point different from the first point. The second tangent line passes through the sensor portion and is tangent to the predicted contour.

11. The object tracking device according to claim 9, wherein: The shape model is a rectangular model having a first vertex, a second vertex, a third vertex, a fourth vertex, a first side, and a second side, wherein the first vertex is farthest from the sensor portion, the second vertex and the third vertex are adjacent to the first vertex, the fourth vertex is located between the second vertex and the third vertex, the first side connects the fourth vertex to the second vertex, and the second side connects the fourth vertex to the third vertex. The prescribed range includes the first side and the second side.

12. The object tracking device according to claim 8, wherein: The predetermined range includes a first range in which the sensor unit can directly irradiate the sensor wave, and a second range in which the sensor unit cannot directly irradiate the sensor wave. The plurality of predicted values ​​include a first predicted value within the first range and a second predicted value within the second range. The estimation unit is configured to have a first contribution degree higher than a second contribution degree, wherein the first contribution degree is a contribution degree of the first predicted value to the update of the estimated value, and the second contribution degree is a contribution degree of the second predicted value to the update of the estimated value.

13. The object tracking device according to claim 1, wherein: The predicted value calculation unit is configured to change the predetermined range according to the distance of the target object.

14. The object tracking device according to claim 13, wherein: The above-mentioned prediction value calculation unit is composed of: When the distance to the target object is greater than or equal to a predetermined distance, the entire periphery of the predicted contour is set as the predetermined range. When the distance to the target object is less than a predetermined distance, the range in which the sensor unit can directly irradiate the sensor wave is set to the predetermined range.

Images