Signal processing apparatus

Abstract

To provide a signal processing apparatus that can improve determination accuracy of a micro-Doppler.SOLUTION: A signal processing apparatus comprises: a feature amount calculation unit 22 that calculates a feature amount associated with monotonously increasing or monotonously decreasing of relative velocity of an observation point for a sensor unit 10 transmitting and receiving an electromagnetic wave to an azimuth direction of the observation point acquired by transmission and reception of the electromagnetic wave; and a determination unit 23 that determines whether or not micro-Doppler is generated in the observation point, on the basis of the feature amount. If the micro-Doppler is generated, since the relative velocity has a tendency to monotonously increase or monotonously decrease to the azimuth direction of the observation point, determination accuracy of the micro-Doppler can be improved by using this tendency.SELECTED DRAWING: Figure 1

Description

[Technical field]

[0001] The present disclosure relates to a signal processing device. [Background technology]

[0002] Object detection devices using millimeter wave radar of the FMCW (Frequency Modulated Continuous Wave) system or LiDAR (Light Detection and Ranging) measure the distance to objects around the device and the relative speed of the objects using the Doppler effect. When such object detection devices are mounted on vehicles and used to recognize the driving environment, the observed information such as distance and relative speed is useful for recognizing state quantities such as the positions and moving speeds of moving objects such as surrounding vehicles and stationary objects such as guardrails.

[0003] However, when a nearby vehicle is moving, the rotating tire measures a relative speed different from that of the body around the tire. This phenomenon is called micro-Doppler and causes a decrease in the accuracy of recognition of vehicle speed and other state quantities.

[0004] The signal processing device described in Patent Document 1 determines whether or not micro-Doppler is occurring using statistics such as the average value and standard deviation of the relative velocity of an observation point, and the difference in relative velocity between surrounding observation points.

[0005] Specifically, the system performs clustering on the observation points with a size limit of about the size of a tire wheel.Then, it calculates the average value and standard deviation of the relative speed for the observation points in the small clusters obtained by clustering, and performs micro-Doppler judgment on each observation point in the small cluster based on the calculation result and the relative speed of the observation points in the vicinity of the small cluster.

[0006] Next, clustering is performed with a size limit of about the width of a vehicle. Then, among the observation points in the obtained medium clusters, those that were determined not to have micro-Doppler in the small cluster determination and whose relative velocities deviate from the average are determined to have micro-Doppler.

[0007] Furthermore, clustering is performed with a size limit of about the length of the vehicle, and from the observation points in the obtained large cluster, observation points that are determined to be generating micro-Doppler in the small and medium clusters are excluded.Then, the speed representative value of the large cluster is calculated using the speed of the remaining observation points. [Prior art documents] [Patent documents]

[0008] [Patent Document 1] JP 2021-124422 A Summary of the Invention [Problem to be solved by the invention]

[0009] Patent Document 1 is based on the premise that observation points are densely obtained from the tires and also from the surrounding body, so there is a risk that the accuracy of micro-Doppler determination will decrease in scenes that do not meet this premise.

[0010] In view of the above, an object of the present disclosure is to provide a signal processing device capable of improving the accuracy of micro-Doppler determination. [Means for solving the problem]

[0011] In order to achieve the above object, according to one aspect of the present disclosure, a signal processing device includes a feature calculation unit (22) that calculates a feature relating to a monotonically increasing or decreasing relative velocity of the observation point with respect to a sensor unit (10) that transmits and receives electromagnetic waves, with respect to the orientation of the observation point obtained by transmitting and receiving electromagnetic waves, and a determination unit (23) that determines whether or not micro-Doppler is occurring at the observation point based on the feature.

[0012] When micro-Doppler occurs, the relative velocity in the direction of the observation point tends to monotonically increase or decrease, and by utilizing this tendency, the accuracy of micro-Doppler determination can be improved.

[0013] The reference symbols in parentheses attached to each component indicate an example of the correspondence between the component and specific components described in the embodiments described below. [Brief description of the drawings]

[0014] [Figure 1] 1 is a block diagram of an object detection device according to a first embodiment. [Diagram 2] FIG. 1 is a diagram for explaining a micro-Doppler. [Diagram 3] 13 is a flowchart of an object detection process. [Figure 4] 13 is a flowchart of a micro-Doppler determination process. [Diagram 5] 13 is a flowchart of a regression process using two straight lines. [Figure 6] 13 is a flowchart of a flat line generation process. [Figure 7] 13 is a flowchart of a slope straight line generation process. [Figure 8] FIG. 13 is a diagram illustrating an example of a distribution of observation points. [Figure 9A] FIG. 13 is a diagram for explaining micro-Doppler determination of a comparative example. [Figure 9B] FIG. 13 is a diagram for explaining micro-Doppler determination of a comparative example. [Figure 9C] FIG. 13 is a diagram for explaining micro-Doppler determination of a comparative example. [Figure 10A] FIG. 2 is a diagram for explaining micro-Doppler determination according to the first embodiment. [Figure 10B] FIG. 2 is a diagram for explaining micro-Doppler determination according to the first embodiment. [Figure 10C] FIG. 2 is a diagram for explaining micro-Doppler determination according to the first embodiment. [Figure 11] 10 is a flowchart of a micro-Doppler determination process in the second embodiment. [Figure 12] 13 is a flowchart of a slope straight line generation process in the third embodiment. [Figure 13] 11 is a diagram for explaining a method for calculating the inclination of a slope line. FIG. [Figure 14] 11 is a diagram for explaining a method for calculating the inclination of a slope line. FIG. [Figure 15] 11 is a diagram for explaining a method for calculating the inclination of a slope line. FIG. [Figure 16] 13 is a flowchart of an object detection process in the fourth embodiment. DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS

[0015] Hereinafter, embodiments of the present disclosure will be described with reference to the drawings. In the following embodiments, parts that are the same or equivalent to each other will be denoted by the same reference numerals.

[0016] (First embodiment) A first embodiment will be described. The object detection device of this embodiment is mounted on a vehicle and recognizes objects around the vehicle. The vehicle on which the object detection device is mounted is referred to as a host vehicle, and a vehicle other than the host vehicle is referred to as a second vehicle. As shown in FIG. 1, the object detection device includes a sensor unit 10 and a signal processing device 20.

[0017] The sensor unit 10 detects an object by transmitting and receiving electromagnetic waves. The sensor unit 10 is composed of, for example, an FMCW radar or LiDAR, and acquires information on a portion of the surface of the object that reflects the electromagnetic waves transmitted by the sensor unit 10 (hereinafter, referred to as an "observation point").

[0018] In this embodiment, the sensor unit 10 scans electromagnetic waves in both the horizontal and vertical directions to obtain the position of the observation point and its relative speed with respect to the sensor unit 10 (hereinafter simply referred to as "relative speed"). Specifically, the position information of the observation point includes the distance between the sensor unit 10 and an object and the direction of the object. The information of the observation point obtained by the sensor unit 10 is transmitted to the signal processing device 20.

[0019] The signal processing device 20 processes the information of the observation point acquired by the sensor unit 10 and recognizes an object. The signal processing device 20 includes a small area generating unit 21, a feature amount calculating unit 22, a determining unit 23, and an object recognizing unit 24.

[0020] The small area generating unit 21 generates small areas for processing in the feature amount calculating unit 22 and the like. Specifically, the small area generating unit 21 sets small areas so as to divide the space around the vehicle into a plurality of areas. For example, the small area generating unit 21 sets a plurality of small areas so as to divide the space around the vehicle into a plurality of areas in each of the vehicle front-rear direction, left-right direction, and up-down direction. Information on the position and size of each small area set by the small area generating unit 21 is transmitted to the feature amount calculating unit 22.

[0021] The feature calculation unit 22 calculates a feature that correlates the direction and speed of each observation point. For each small area generated by the small area generation unit 21, the feature calculation unit 22 calculates a feature that correlates the direction and speed of the observation point included in the small area. In detail, the feature calculation unit 22 calculates a feature related to the monotonous increase or decrease of the relative speed with respect to the direction of the observation point.

[0022] In this embodiment, the feature calculation unit 22 regresses the data of the azimuth and relative speed of the observation point on a slope line where the relative speed increases or decreases monotonically with respect to the azimuth, and a flat line where the relative speed is constant, and outputs the result of this regression as a feature. In this embodiment, the feature calculation unit 22 calculates the feature using a vertical angle as the azimuth of the observation point. The vertical angle here is the angle between a line connecting the sensor unit 10 and the observation point and a horizontal plane passing through the sensor unit 10.

[0023] When the sensor unit 10 detects the rotating tire of another vehicle, a micro-Doppler is generated at the observation point of the tire, in which the relative speed is reflected not only in the ground speed of the other vehicle but also in the tire rotational speed, and a relative speed different from that measured at the observation point on the body around the tire is measured.

[0024] For example, when the tire of another vehicle is rotating as shown by arrow A1 in Fig. 2, the observation points on the upper half of the tire have a speed moving away from the sensor unit 10, and the observation points on the lower half of the tire have a speed moving toward the sensor unit 10. These velocities increase the further away from the center of the tire. Therefore, the relative velocity of the observation points on the tire increases monotonically with respect to the vertical angle. Also, when the tire is rotating in the opposite direction to the arrow A1, the relative velocity of the observation points on the tire decreases monotonically with respect to the vertical angle.

[0025] The feature amount calculation unit 22 captures the change in relative speed with respect to the vertical angle for the observation points of the tires as a slope line, whereas the feature amount calculation unit 22 captures the change in relative speed with respect to the vertical angle for the observation points of the surrounding body as a flat line where the relative speed is constant.

[0026] In addition, when a sensor such as a radar or LiDAR is mounted on a vehicle, the mounting position is often the front grille or the upper part of the windshield of the vehicle. The relative speed of the observation point of the surrounding body of another vehicle changes strictly according to the cosine of the vertical angle, but when the sensor unit 10 is mounted in such a place, the height of the observation point relative to the sensor unit 10 is often sufficiently small compared to the distance between the sensor unit 10 and the observation point. Therefore, the vertical angle becomes a small value, and the change in value due to the cosine of the vertical angle is almost 0. Therefore, even if it is assumed that the relative speed of the observation point of the surrounding body is constant, it is considered that the effect of this embodiment described later can be obtained.

[0027] The determination unit 23 determines whether or not micro-Doppler occurs at each observation point based on the feature calculated by the feature calculation unit 22. Specifically, the determination unit 23 determines that, among the observation points included in the small area to be processed, micro-Doppler occurs at the observation points belonging to the slope line, and determines that micro-Doppler does not occur at the observation points belonging to the flat line. The determination result by the determination unit 23 is transmitted to the object recognition unit 24.

[0028] The object recognition unit 24 recognizes objects around the vehicle based on the information on the observation points acquired by the sensor unit 10 and the judgment results of the judgment unit 23. The object recognition unit 24 generates a dynamic occupancy lattice map, generates a probability distribution of speed in each cell of the dynamic occupancy lattice map, and performs observation and updating to recognize the positions and movement states of objects around the vehicle.

[0029] When setting the initial value of the probability distribution of the speed, the object recognition unit 24 excludes the observation points determined by the determination unit 23 to be generating micro-Doppler, and sets the initial value using the average value, variance value, etc. of the relative speeds of the remaining observation points. The object recognition unit 24 also excludes the observation points determined by the determination unit 23 to be generating micro-Doppler, and performs likelihood calculations when updating the distribution using the relative speeds of the remaining observation points.

[0030] The object recognition result by the object recognition unit 24 is transmitted to an ECU (Electronic Control Unit) (not shown) and used in ADAS (Advanced driver-assistance systems) applications, autonomous driving applications, fusion systems with other sensors, and the like.

[0031] The operation of the object detection device will now be described. The object detection device recognizes an object by sequentially executing steps S101 to S104 shown in FIG.

[0032] In step S101, the object detection device acquires an observation point and performs pre-processing of step S102. Specifically, the signal processing device 20 acquires the position and relative speed of the observation point detected by the sensor unit 10, and performs removal of noise points that are not used for object recognition, etc.

[0033] In the next step S102, the signal processing device 20 performs a micro-Doppler determination. That is, the signal processing device 20 determines whether or not a micro-Doppler occurs at each observation point acquired in step S101. Details of step S102 will be described later.

[0034] In the next step S103, the signal processing device 20 performs DOGM (Dynamic Occupancy Grid Mapping). That is, the object recognition unit 24 generates a dynamic occupancy grid map and recognizes objects around the vehicle. In step S103, the object recognition unit 24 generates the dynamic occupancy grid map by excluding the observation point where it is determined that micro-Doppler is occurring.

[0035] In the following step S104, the signal processing device 20 outputs the recognition result in step S103 to an ECU (not shown) etc. After step S104, the process ends.

[0036] Step S102 will now be described in detail. In step S102, the process shown in FIG. 4 is executed. First, in step S201, the small area generating unit 21 generates small areas. Specifically, the small area generating unit 21 divides the space around the vehicle into a plurality of small areas in each of the forward / backward direction, left / right direction, and up / down direction, and sets the plurality of small areas, and outputs the position and size of each small area. The number of generated small areas is set to N1, and i1 is an integer between 1 and N1, and the i1th small area is set as small area C. i1 Let us assume that.

[0037] After step S201, the signal processing device 20 calculates the small areas C1 to C N1 A loop process starting from step S202 and ending in any one of steps S205 to S207 is executed for the above.

[0038] In step S202, the small area generating unit 21 generates a small area C i1 Specifically, the small area generating unit 21 acquires data of observation points included in the small area C i1 The vertical angle and relative velocity of the observation points included in the small area C i1 If there are multiple observation points included in the region, the small region generating unit 21 obtains the vertical angle and relative velocity of each of the multiple observation points.

[0039] In the next step S203, the feature amount calculation unit 22 executes a regression process using two straight lines on the data of the observation points. The details of step S203 will be described later.

[0040] In the next step S204, the determination unit 23 determines whether or not a flat line has been established. Specifically, the determination unit 23 determines whether or not an establishment flag of the flat line is TRUE.

[0041] If it is determined that the flat line establishment flag is TRUE, the process proceeds to step S205. In step S205, the determination unit 23 determines that micro-Doppler is occurring at an observation point that does not belong to the flat line. The observation point that does not belong to the flat line is an observation point that has been excluded from a set F, which will be described later. After step S205, the current loop process ends.

[0042] If it is determined that the establishment flag of the flat line is not TRUE, the process proceeds to step S206. In step S206, the determination unit 23 determines whether or not the establishment flag of the slope line is TRUE. Specifically, the determination unit 23 determines whether or not the establishment flag of the slope line is TRUE.

[0043] If it is determined that the establishment flag of the slope line is TRUE, the process proceeds to step S207. In step S207, the determination unit 23 determines that micro-Doppler is occurring at the observation points belonging to the slope line. The observation points belonging to the slope line are the observation points included in the set S described below. After step S207, the current loop process ends. If it is determined that the establishment flag of the slope line is not TRUE, the current loop process ends. Small regions C1 to C N1 When the loop process is completed for all of the above, the process of step S102 ends.

[0044] Details of step S203 will be described. In step S203, the process shown in Fig. 5 is performed. In step S301, the feature calculation unit 22 generates a flat line. In the following step S302, the feature calculation unit 22 generates a slope line. After step S302, the process of step S203 ends.

[0045] Details of step S301 will be described with reference to FIG. 6. In step S401, the feature amount calculation unit 22 calculates the feature amount of the small region C i1 Determine whether there are any observation points included in small area C. i1If it is determined that there is no observation point included in the range, the process of step S301 ends, and if it is determined that there is an observation point included in the range, the process proceeds to step S402.

[0046] The feature amount calculation unit 22 performs linear regression on the data of the observation points in descending order of azimuth. Specifically, in step S402, the feature amount calculation unit 22 performs linear regression on the data of the observation points in descending order of azimuth. i1 The observation points in small area C are sorted in descending order of vertical angle. i1 Let N2 be the number of observation points included in, and let i2 be an integer between 1 and N2. Among the sorted observation points, the i2-th observation point from the one with the largest vertical angle is called observation point P. i2 In addition, the feature amount calculation unit 22 sets and initializes a set F of observation points. In the initial state, the number of observation points included in the set F is 0. After step S402, the feature amount calculation unit 22 sets and initializes the set F of observation points P1 to P N2 A loop process starting from step S403 and ending in step S405 or step S406 is executed for the above.

[0047] In step S403, the feature amount calculation unit 22 calculates the observation point P i2 to set F. In the following step S404, the feature calculation unit 22 calculates the average value of the relative velocities of the observation points included in set F and the RMSE (Root Mean Squared Error). A straight line along which the relative velocity is constant at this calculated average value becomes a flat line. In the following step S405, the feature calculation unit 22 determines whether the RMSE calculated in step S404 is equal to or smaller than a predetermined threshold value.

[0048] If it is determined in step S405 that the RMSE is equal to or smaller than the threshold, the current loop process ends. If it is determined in step S405 that the RMSE is not equal to or smaller than the threshold, the process proceeds to step S406. In step S406, the feature amount calculation unit 22 calculates the feature amount of the observation point P i2 is excluded from the set F. After step S406, the current loop process ends. N2 When the loop process is completed for all of the above, the process proceeds to step S407.

[0049] In step S407, feature amount calculation unit 22 determines whether the number of observation points included in set F is equal to or greater than a predetermined threshold. If it is determined that the number of observation points included in set F is equal to or greater than the threshold, the process proceeds to step S408, where feature amount calculation unit 22 sets the establishment flag of the flat line to TRUE. After step S408, the process of step S301 ends.

[0050] If it is determined that the number of observation points included in the set F is not equal to or greater than the threshold, the process proceeds to step S409, where the feature calculation unit 22 sets the establishment flag of the flat line to FALSE. After step S409, the process of step S301 ends.

[0051] Details of step S302 will be described with reference to Fig. 7. In step S501, the feature calculation unit 22 determines whether or not a flat line is established. Specifically, the feature calculation unit 22 determines whether or not a flat line establishment flag is TRUE. In addition, the feature calculation unit 22 sets and initializes a set S of observation points. In the initial state, the number of observation points included in the set S is 0.

[0052] If it is determined that the flat line establishment flag is TRUE, the process proceeds to step S502. N2 Among them, the observation points excluded from set F are added to set S. After step S502, the process proceeds to step S504.

[0053] If it is determined that the flat line establishment flag is not TRUE, the process proceeds to step S503. N2 is added to the set S. After step S503, the process proceeds to step S504.

[0054] In step S504, the feature calculation unit 22 calculates a regression line of the set S. This regression line becomes a slope line. Specifically, the feature calculation unit 22 calculates an approximation line of the vertical angle and the relative speed for the observation points included in the set S using the least squares method. Note that approximation using a Hough transform or a piecewise linear function may also be used. The slope of the regression line of the set S is defined as M1.

[0055] In the following step S505, feature calculation unit 22 determines whether or not the number of observation points included in set S is equal to or greater than a predetermined threshold and whether or not gradient M1 is equal to or greater than a predetermined threshold.

[0056] If it is determined that the number of observation points included in the set S is equal to or greater than the threshold and that the slope M1 is equal to or greater than the threshold, the process proceeds to step S506. In step S506, the feature calculation unit 22 sets the establishment flag of the slope line to TRUE. After step S506, the process of step S302 ends.

[0057] If it is determined that the number of observation points included in the set S is not equal to or greater than the threshold, or that the slope M1 is not equal to or greater than the threshold, the process proceeds to step S507. In step S507, the feature calculation unit 22 sets the establishment flag of the slope line to FALSE. After step S507, the process of step S302 ends.

[0058] The effect of this embodiment will be described. Here, as shown in Fig. 8, a scene after another vehicle 200 has overtaken the host vehicle 100 will be described. An arrow A2 indicates the traveling direction of the other vehicle 200. An area R1 shown in Fig. 8 is an area around the rear tire of the other vehicle 200 on the side closer to the host vehicle 100. An area R2 is an area around the front tire. An area R3 is an area around the rear tire on the side farther from the host vehicle 100.

[0059] In region R1, observation points are easily obtained densely on the tire and also on the surrounding body. In region R2, observation points are easily obtained densely on the tire, but are difficult to obtain on the surrounding body. In region R3, observation points are easily obtained only on the lower half of the tire, and are easy to obtain on the surrounding body.

[0060] For example, let us assume that micro-Doppler determination is performed by calculating the average value of the relative velocities of the observation points included in a small region and comparing this average value with the relative velocities of each observation point. In this case, for region R1, as shown in Fig. 9A, the difference between the average value of the relative velocities of the observation points of the surrounding body is small, so it is correctly determined that micro-Doppler is not occurring. Also, for the observation points of the tires, the difference between the relative velocities of the data in the dashed line area and the average value is large, so it is correctly determined that micro-Doppler is occurring.

[0061] On the other hand, for region R2, if the observation points of the surrounding bodies are not acquired as shown in Fig. 9B, it is not possible to compare the average relative speed with the average relative speed of the observation points of the surrounding bodies, and therefore a situation may arise in which the judgment cannot be performed. Also, for region R3, as shown in Fig. 9C, a discrepancy occurs between the average relative speed and the relative speed of the surrounding bodies, so it is judged that micro-Doppler is occurring at most of the observation points in the region, and an erroneous judgment is made for the observation points of the surrounding bodies.

[0062] In this embodiment, for region R1, as shown in FIG. 10A, the observation points of the surrounding body belong to a flat straight line, and the observation points of the tires surrounded by dashed lines belong to a slope straight line, and it is correctly determined that micro-Doppler is occurring.

[0063] For region R2, if no observation points of the surrounding body are acquired as shown in Fig. 10B, the flat line does not hold, but in step S503, all observation points of region R2 are added to set S and a slope line is calculated. As a result, it is correctly determined that micro-Doppler is occurring at the observation points surrounded by the dashed line.

[0064] For region R3, even if only the observation points of the lower half of the tire are acquired as shown in Fig. 10C, the observation points of the surrounding body are added to set F and the observation points of the tire are added to set S based on the relationship between the vertical angle and the relative velocity. This allows a correct determination that micro-Doppler is occurring at the observation points surrounded by the dashed line.

[0065] In this way, the range of application of micro-Doppler determination can be expanded by using the linear regression results on the two-dimensional coordinate axis of the vertical angle and the relative velocity, which is a higher-order feature value than statistics such as the average value of the relative velocity in a small region. This embodiment, which uses the vertical angle as the azimuth of the observation point, is particularly suitable for LiDAR, where observation points are expected to be obtained densely in the vertical direction.

[0066] When generating the initial velocity distribution in step S103, if the observation points used for generation include observation points where micro-Doppler occurs, an error may occur between the calculated average value and the relative velocity of the body part, or the variance value may become unnecessarily large. If such an error occurs, the result of generating the initial velocity distribution may have a high probability of being in a position that is out of the relative velocity of the body part, which may affect the velocity estimation accuracy and convergence speed. In contrast, by performing micro-Doppler judgment as in this embodiment and removing the observation points where it is judged that micro-Doppler occurs, the error and variance value between the calculated average value of the relative velocity and the relative velocity of the body part are reduced, and the velocity estimation accuracy, etc. are improved.

[0067] Furthermore, when updating the probability distribution of the speed, the likelihood of a speed close to the relative speed of the observation point included in each cell is usually calculated to be high. Therefore, if the observation points used for updating include observation points where micro-Doppler occurs, the likelihood of a speed that deviates from the speed of the body part is calculated to be high, which may affect the speed estimation accuracy and convergence speed. In contrast, by performing micro-Doppler judgment as in this embodiment and removing the observation points where it is judged that micro-Doppler occurs, only observation points close to the speed of the body part are used in the likelihood update, improving the speed estimation accuracy, etc.

[0068] The inventors conducted an experiment using an object detection device that does not perform micro-Doppler determination and the object detection device of this embodiment, and performed velocity estimation using a dynamic occupancy grid map in the scene shown in Fig. 8. The results of this experiment showed that, compared to the object detection device that does not perform micro-Doppler determination, the object detection device of this embodiment had reduced estimation error and estimation variance, and in particular, the estimation variance was significantly reduced.

[0069] As described above, in this embodiment, the accuracy of micro-Doppler determination can be improved by utilizing the relationship in which the relative velocity monotonically increases or decreases with respect to the azimuth of the observation point.

[0070] Furthermore, according to the above embodiment, the following effects can be obtained.

[0071] (1) The feature amount calculation unit 22 regresses the observation points onto a flat line in descending order of azimuth, and regresses the observation points whose error amount from the flat line is greater than a threshold onto a slope line. The larger the vertical angle, i.e., the greater the distance from the ground, the more likely the observation point is a body part. Therefore, by adding the observation points to the set F in descending order of vertical angle and calculating the flat line, the flat line will more appropriately capture the body part, improving the accuracy of the micro-Doppler determination.

[0072] (2) The object recognition unit 24 recognizes objects using a dynamic occupancy grid map generated by excluding observation points determined to be generating micro-Doppler by the determination unit 23. This improves the accuracy of speed estimation, etc.

[0073] Second embodiment The second embodiment will be described. This embodiment is different from the first embodiment in the processing when a flat line is established, and other aspects are the same as the first embodiment, so only the differences from the first embodiment will be described.

[0074] As shown in Fig. 11, in this embodiment, the signal processing device 20 executes step S208 instead of step S205. In step S208, the signal processing device 20 replaces the relative velocities of the observation points not belonging to the flat line with the average value of the relative velocities of the observation points belonging to the flat line. Specifically, the determination unit 23 sets the average value of the relative velocities of the observation points included in the set F as V, and calculates the average value of the relative velocities of the observation points P1 to P N2 For observation points excluded from set F, the relative velocity is replaced with relative velocity V, and it is determined that micro-Doppler is not occurring. For observation points not belonging to the flat line, the object recognition unit 24 performs calculations of the probability distribution of the velocity using the replaced relative velocity.

[0075] When a flat line capturing the surrounding body is established, the speed of the other vehicle can be recognized with a certain degree of accuracy, and it is considered that even if the relative speed of the observation point excluded from the set F is replaced as described above, the influence on the recognition of the speed of the other vehicle is small. By replacing the relative speed in this way and using it for the calculation in the object recognition unit 24, it is possible to suppress a reduction in the number of observation points used for the calculation in the object recognition unit 24. In this embodiment as well, the speed estimation accuracy and the like are improved, as in the first embodiment.

[0076] This embodiment has the same configuration and operation as the first embodiment, and therefore can obtain the same effects as the first embodiment.

[0077] Furthermore, according to the above embodiment, the following effects can be obtained.

[0078] (1) When the feature amount calculation unit 22 determines that the flat line is valid, the determination unit 23 replaces the relative velocity with the relative velocity V for the observation points excluded from the flat line and determines that micro-Doppler is not occurring. This makes it possible to suppress a reduction in the number of observation points used for calculations in the object recognition unit 24.

[0079] Third embodiment A third embodiment will be described. This embodiment is the same as the first embodiment except that the method of generating the slope line is changed. Therefore, only the differences from the first embodiment will be described.

[0080] In step S302, the feature calculation unit 22 of this embodiment executes the process shown in Fig. 12 instead of the process shown in Fig. 7. First, in step S601, the feature calculation unit 22 determines whether or not a flat line has been established, similarly to step S501, and sets a set S. If it is determined that a flat line has been established, the process proceeds to step S602.

[0081] In step S602, the feature amount calculation unit 22 calculates the gradient of the slope line. The gradient calculated here is designated as M2. A method for calculating the gradient M2 will be described later. After step S602, the feature amount calculation unit 22 calculates the gradient of the slope line from the observation points P1 to P N2 A loop process starting from step S603 and ending in either step S603, step S606, or step S607 is executed for the above.

[0082] In step S603, the feature amount calculation unit 22 calculates the observation point P i2 Determine whether the observation point P belongs to the set F. i2 If it is determined that the observation point P belongs to the set F, the loop processing ends. i2 If it is determined that does not belong to set F, the process proceeds to step S604.

[0083] In step S604, the feature amount calculation unit 22 i2 to the set S. In the following step S605, the feature calculation unit 22 calculates a regression line of the set S using the slope M2 as a constraint condition. In addition, the feature calculation unit 22 calculates the RMSE.

[0084] In the next step S606, the feature calculation unit 22 determines whether the RMSE calculated in step S605 is equal to or less than a predetermined threshold. If it is determined that the RMSE is equal to or less than the threshold, the current loop process ends. If it is determined that the RMSE is not equal to or less than the threshold, the process proceeds to step S607.

[0085] In step S607, the feature amount calculation unit 22 i2 is excluded from the set S. After step S607, the current loop process ends. N2 When the loop process has been completed for all of the above, the process proceeds to step S608.

[0086] In step S608, the feature calculation unit 22 determines whether the number of observation points included in the set S is equal to or greater than a predetermined threshold. If it is determined that the number of observation points included in the set S is equal to or greater than the threshold, the process proceeds to step S609. In step S609, the feature calculation unit 22 sets the establishment flag of the slope straight line to TRUE. After step S609, the process of step S302 ends.

[0087] If it is determined that the number of observation points included in the set S is not equal to or greater than the threshold, the process proceeds to step S610. In step S610, the feature calculation unit 22 sets the establishment flag of the slope line to FALSE. After step S610, the process of step S302 ends.

[0088] If it is determined in step S601 that the flat line does not exist, the process proceeds to step S611, and the feature calculation unit 22 executes the same processes as steps S503 to S507 in FIG.

[0089] That is, in step S611, the feature amount calculation unit 22 calculates the feature amount of the observation points P1 to P N2 to set S. In the following step S612, feature calculation unit 22 calculates a regression line of set S. In the following step S613, feature calculation unit 22 determines whether or not the number of observation points included in set S is equal to or greater than a predetermined threshold and whether or not the slope M1 of the regression line calculated in step S612 is equal to or greater than a predetermined threshold.

[0090] If it is determined that the number of observation points included in set S is equal to or greater than the threshold and that the slope M1 is equal to or greater than the threshold, the process proceeds to step S609. If it is determined that the number of observation points included in set S is not equal to or greater than the threshold or that the slope M1 is not equal to or greater than the threshold, the process proceeds to step S610.

[0091] Details of step S602 will be described. In step S602, the feature calculation unit 22 estimates the gradient M2 of the slope line from the ground speed of the vehicle, the tire diameter of the other vehicle, the mounting position of the sensor unit 10, etc., and calculates the slope line using this gradient M2. The gradient M2 can be calculated, for example, as follows.

[0092] As shown in FIG. 13, the small area to be processed is designated as C. The small area C includes the tires 210 and the body of the other vehicle 200. If the ground speed of the host vehicle is V e The direction of the small area C with respect to the traveling direction of the vehicle in the horizontal plane is ω. The ground speed of the other vehicle 200 is V t The direction of the small area C with respect to the traveling direction of the other vehicle 200 in the horizontal plane is τ, and the relative speed of the body of the other vehicle 200 in the direction τ is V r Then, V r =V t cosτ-V e It becomes cosω.

[0093] As shown in FIG. 14, the relative speed of the upper end of the tire 210 is V rtop Then, V rtop =2V t cosτ-V ecosω=2(V t cosτ-V e cosω)+V e cosω=2V r +V e cosω. Also, the relative velocity of the lower end of the tire 210 is V rbtm Then, V rbtm =-V e It becomes cosω.

[0094] If the vertical angle of the upper end of the tire 210 is φ, the relative velocity of the observation point of the upper end of the tire 210 is V rtop cosφ. If the vertical angle of the bottom end of the tire 210 is θ, the relative velocity of the observation point of the bottom end of the tire 210 is V rbtm The result is cosθ.

[0095] 15, on a two-dimensional coordinate axis of vertical angle and relative velocity, the plots of the observation points of the tire 210 are expected to be aligned on a straight line connecting the plots of the upper end and the lower end of the tire 210. Strictly speaking, the relative velocity varies nonlinearly with respect to the vertical angle, but as described above, when the vertical angle is a small value, the relative velocity varies approximately linearly. Therefore, the ground speed V e , relative velocity V r If the direction ω, angles φ, and θ are obtained, the gradient M2 of the slope line can be estimated.

[0096] These parameters can be obtained, for example, as follows: e The relative velocity V can be measured by a wheel speed sensor, an IMU (Inertial Measurement Unit), a GPS (Global Positioning System), or other ego-motion sensor mounted on the vehicle. r For this, it is possible to use the speed recognized on the flat line of the small area C. The direction ω can be calculated from the position of the small area C set in step S201.

[0097] If the distance from the sensor unit 10 to the small area C is L, the diameter of the tire 210 is R, and the mounting height of the sensor unit 10 is Z, the angles φ and θ can be calculated from the distance L, the diameter R, and the mounting height Z. The distance L can be calculated from the position of the small area C set in step S201. The average tire diameter of an automobile can be used as the diameter R. Alternatively, a plurality of patterns of the diameter R may be set, and the diameter R may be selected from the plurality of patterns according to the size of the other vehicle 200 acquired by the processing in the object recognition unit 24. The sensor mounting height Z can be a value stored in advance as a set value.

[0098] In step S605, the feature calculation unit 22 calculates a regression line using the slope M2 calculated using these parameters as a constraint condition.

[0099] By using information on the estimated slope M2 as a constraint condition during linear regression, it is possible to prevent points where no micro-Doppler occurs from belonging to the slope line, thereby improving the accuracy of micro-Doppler determination.

[0100] This embodiment has the same configuration and operation as the first embodiment, and therefore can obtain the same effects as the first embodiment.

[0101] Furthermore, according to the above embodiment, the following effects can be obtained.

[0102] (1) The feature amount calculation unit 22 calculates the gradient M2 of the slope line based on the ground speed of the vehicle, the tire diameter of the other vehicle, and the mounting position of the sensor unit 10, and generates the slope line with the calculated gradient M2. This can improve the accuracy of the micro-Doppler determination.

[0103] (Fourth embodiment) A fourth embodiment will be described. In this embodiment, the process executed by the object recognition unit 24 in the first embodiment is changed, and the rest is the same as in the first embodiment, so only the parts that are different from the first embodiment will be described.

[0104] As shown in FIG. 16, the signal processing device 20 of this embodiment executes steps S105 and S106 instead of step S103 following step S102.

[0105] In step S105, the object recognition unit 24 performs clustering. Clustering is a process of grouping together a group of observation points that are presumed to be observation points of the same object, and calculating the shape, size, and the like of the object.

[0106] The object recognition unit 24 excludes the observation points determined to be generating micro-Doppler in step S102, and groups the remaining observation points that are presumed to be observation points of the same object. Then, the object recognition unit 24 calculates the shape, size, etc. of each of the generated groups.

[0107] In the next step S106, the object recognition unit 24 performs tracking. That is, for the groups generated in step S105, the object recognition unit 24 calculates state quantities as targets, such as positions and velocities, from the displacements of those that can be regarded as the same object in time series. After step S106, the process proceeds to step S104.

[0108] When clustering is performed including an observation point where micro-Doppler is occurring, this observation point may be divided into a different group from other observation points, which may result in erroneous recognition or over-division of targets, etc. In this case, there is a risk that the error of the average velocity value from the true velocity may become large during tracking, or the variance value may become unnecessarily large, thereby increasing the error and variation in the velocity accuracy of the target.

[0109] In response to this, by performing clustering while excluding observation points determined to be generating micro-Doppler, it is possible to suppress erroneous recognition and over-division of targets, etc. Also, it is possible to reduce the error and variance of the average velocity value relative to the true value, thereby reducing the error and variation in the velocity accuracy of targets.

[0110] This embodiment has the same configuration and operation as the first embodiment, and therefore can obtain the same effects as the first embodiment.

[0111] Furthermore, according to the above embodiment, the following effects can be obtained.

[0112] (1) The object recognition unit 24 performs clustering by excluding observation points determined by the determination unit 23 to be generating micro-Doppler. This makes it possible to suppress erroneous recognition and over-division of targets. In addition, it is possible to reduce errors and variations in the speed accuracy of targets.

[0113] (Other embodiments) The present disclosure is not limited to the above-described embodiments, and can be modified as appropriate. The above-described embodiments are not unrelated to each other, and can be combined as appropriate, except in cases where the combination is clearly impossible. In addition, in the above-described embodiments, it goes without saying that the elements constituting the embodiments are not necessarily essential, except in cases where they are particularly clearly specified as essential or where they are clearly considered essential in principle.

[0114] In the second embodiment, the slope M2 may be estimated and linear regression may be performed as in the third embodiment. In the second embodiment, clustering and tracking may be performed as in the fourth embodiment. By replacing the relative speed of the observation point excluded from the set F with the relative speed V and reducing the speed difference with the observation point of the surrounding body, the tire and the surrounding body are more likely to be classified into the same group. Therefore, it is possible to suppress erroneous recognition and over-division of the target, and reduce errors and variations in the speed accuracy of the target. In the third embodiment, clustering and tracking may be performed as in the fourth embodiment.

[0115] In each of the above embodiments, the feature amount calculation unit 22 regresses the data of the observation point onto the slope line and the flat line, and outputs the regression result as a feature amount. However, the feature amount calculation unit 22 may calculate another feature amount related to the monotonous increase or decrease of the relative speed with respect to the azimuth of the observation point.

[0116] The object recognition unit 24 may be configured integrally with the small area generation unit 21, the feature amount calculation unit 22, and the determination unit 23, or may be configured separately. For example, the signal processing device 20 is configured by a microcomputer equipped with a CPU, ROM, RAM, I / O, etc., but the object recognition unit 24 may be configured by a microcomputer separate from the small area generation unit 21, etc. Furthermore, the signal processing device 20 does not need to include the object recognition unit 24. Furthermore, the determination result of the determination unit 23 may be used for purposes other than object recognition. [Explanation of symbols]

[0117] 22 Feature Calculation Unit 23 Judgment section

Claims

1. 1. A signal processing device, comprising: a feature calculation unit (22) that calculates a feature relating to a monotonous increase or decrease in a relative speed of an observation point with respect to a sensor unit (10) that transmits and receives electromagnetic waves, with respect to an orientation of the observation point acquired by transmitting and receiving the electromagnetic waves; A signal processing device comprising: a determination unit (23) that determines whether or not micro Doppler is occurring at the observation point based on the feature amount.

2. The feature amount calculation unit is The azimuth of the observation point and the relative speed with respect to the sensor unit, a slope line along which the relative velocity increases or decreases monotonically with respect to the direction; The relative velocity regresses to a constant flat line, The signal processing apparatus according to claim 1 , wherein the result of the regression is output as the feature quantity.

3. The feature amount calculation unit is determining whether the slope line is established based on the number of the observation points belonging to the slope line and the inclination of the slope line; determining whether the flat line is established based on the number of observation points belonging to the flat line; The signal processing device according to claim 2 , wherein a result of a determination as to whether or not the slope line and the flat line are established is included in the feature amount and is output.

4. The determination unit is When the feature amount calculation unit determines that the flat line is established, it determines that micro Doppler occurs at the observation point excluded from the flat line, The signal processing device according to claim 3 , wherein, when the feature amount calculation unit determines that the slope line is established, it is determined that micro-Doppler is occurring at the observation point belonging to the slope line.

5. The determination unit is When the feature amount calculation unit determines that the flat line is established, Let V be the relative velocity of the observation point belonging to the flat line, The signal processing device according to claim 3 , wherein the relative velocity of the observation point excluded from the flat line is replaced with the relative velocity V, and it is determined that micro-Doppler is not occurring.

6. The feature amount calculation unit is Regressing the observation points onto the flat line in descending order of azimuth; The signal processing device according to claim 2 , wherein the observation points whose error amount from the flat line is greater than a threshold value are regressed onto the slope line.

7. The sensor unit is mounted on a vehicle, A vehicle equipped with the sensor unit is defined as a host vehicle, A vehicle other than the host vehicle is defined as another vehicle. The sensor unit acquires information on the observation point of the other vehicle, 3. The signal processing device according to claim 2, wherein the feature calculation unit calculates a gradient of the slope straight line based on a ground speed of the host vehicle, a tire diameter of the other vehicle, and a mounting position of the sensor unit, and generates the slope straight line based on the calculated gradient.

8. An object recognition unit (24) that recognizes an object based on the information of the observation point, The signal processing device according to claim 1 , wherein the object recognition unit recognizes objects using a dynamic occupancy grid map that is generated by excluding the observation points at which the determination unit determines that micro-Doppler is occurring.

9. An object recognition unit (24) that recognizes an object based on the information of the observation point, The signal processing device according to claim 1 , wherein the object recognition unit performs clustering by excluding the observation points at which the determination unit has determined that micro-Doppler is occurring.

Images