Object detection device and object detection method

Abstract

To reduce processing load in object detection.SOLUTION: An object detection device comprises: a point group data acquisition part acquiring point group data of an object in a radiation range generated by a laser sensor; and an object detection part detecting an object present in the radiation range based on the point group data. The object detection part includes: an area setting part setting a first monitoring area and a second monitoring area positioned above the first monitoring area in the radiation range; a filtering part performing ground filter processing for removing a point corresponding to the ground from a first point group included in the first monitoring area of the point group data; and a clustering part clustering the point group subjected to ground filter processing to detect an object in the first monitoring area, and not performing ground filter processing on a second point group included in the second monitoring area of the point group data but clustering the second point group to detect an object in the second monitoring area.SELECTED DRAWING: Figure 3

Description

[Technical Field]

[0001] The present disclosure relates to an object detection device and an object detection method. [Background technology]

[0002] Patent Document 1 describes a point cloud data collection system that acquires point cloud data including measurement data at multiple measurement points measured by a laser sensor, and when the height of a plane fitted to the measurement points calculated based on the acquired point cloud data is equal to or less than a predetermined height threshold, removes the measurement data of the measurement points that make up the plane as measurement data of measurement points on the ground, and transmits the point cloud data after the measurement data has been removed to a server. [Prior art documents] [Patent documents]

[0003] [Patent Document 1] Patent Publication No. 2021-043475 Summary of the Invention [Problem to be solved by the invention]

[0004] As described above, object detection using a laser sensor generally involves performing ground filtering to remove points corresponding to the ground from the point cloud data measured by the laser sensor, and then analyzing the remaining point cloud to detect objects. Since ground filtering is performed on a large number of point clouds, it requires a large amount of calculation.

[0005] Therefore, an object of the present disclosure is to reduce the processing load of object detection. [Means for solving the problem]

[0006] An object detection device according to one aspect emits laser light toward an illumination area around a vehicle and detects objects present in the illumination area using a laser sensor that receives reflected light of the laser light. The object detection device includes: a point cloud data acquisition unit that acquires point cloud data of objects in the illumination area generated by the laser sensor; and an object detection unit that detects objects present in the illumination area based on the point cloud data. The object detection unit includes: an area setting unit that sets a first monitoring area within the illumination area and a second monitoring area located above the first monitoring area; a filtering unit that performs ground filtering on a first point cloud included in the first monitoring area from the point cloud data to remove points corresponding to the ground; and a clustering unit that performs clustering on the point cloud after the ground filtering to detect objects in the first monitoring area, and performs clustering on a second point cloud included in the second monitoring area from the point cloud data without performing ground filtering to detect objects in the second monitoring area.

[0007] In the object detection device according to the above aspect, a first monitoring area and a second monitoring area are set within the illumination area, and ground filtering is performed on the first point cloud included in the first monitoring area, but not on the second point cloud included in the second monitoring area. In other words, in this object detection device, the area in which ground filtering is performed is limited to the first monitoring area, thereby reducing the processing load. Furthermore, since the second monitoring area is located above the first monitoring area, it is highly likely that the second monitoring area does not include a point cloud corresponding to the ground. Therefore, even when the area in which ground filtering is performed is limited, the point cloud corresponding to the ground can be appropriately removed from the point cloud data.

[0008] The area setting unit may set the first monitoring area and the second monitoring area so that the width of the second monitoring area in the horizontal direction is smaller than the width of the first monitoring area in the horizontal direction. Objects present on the ground may be moving objects such as pedestrians, bicycles, and automobiles. By relatively increasing the width of the first monitoring area in the horizontal direction, it is possible to detect moving objects that may jump out. On the other hand, objects present at a position higher than the ground are likely to be stationary objects such as structures. By relatively decreasing the width of the second monitoring area in the horizontal direction, it is possible to reduce the processing load while appropriately detecting stationary objects that may come into contact with the vehicle.

[0009] The area setting unit may set the first monitoring area and the second monitoring area so that the first monitoring area and the second monitoring area partially overlap each other, in which case the object can be detected appropriately.

[0010] The laser sensor may be a LiDAR, which can detect objects with high accuracy.

[0011] An object detection method according to one aspect includes irradiating an illumination area around a vehicle with laser light and detecting an object present in the illumination area using a laser sensor that receives reflected light of the laser light. The object detection method includes the steps of acquiring point cloud data of the object present in the illumination area generated by the laser sensor and detecting the object present in the illumination area based on the point cloud data, wherein the object detection step includes the steps of setting a first monitoring area and a second monitoring area located above the first monitoring area within the illumination area, performing ground filtering on a first point cloud included in the first monitoring area from the point cloud data to remove points corresponding to the ground, and detecting the object in the first monitoring area by performing clustering processing on the point cloud after the ground filtering processing, and performing clustering processing on a second point cloud included in the second monitoring area from the point cloud data without performing ground filtering on the second point cloud.

[0012] As described above, this object detection method can reduce the processing load. [Effects of the Invention]

[0013] According to various aspects of the present disclosure, the processing load of object detection can be reduced. [Brief explanation of the drawings]

[0014] [Figure 1] 1 is a side view showing a vehicle equipped with an object detection device according to an embodiment. [Figure 2] FIG. 2 is a block diagram showing a functional configuration of the object detection device. [Figure 3] FIG. 2 is a diagram schematically illustrating a first monitoring area and a second monitoring area. [Figure 4] 1 is a flowchart illustrating an object detection method according to an embodiment. DETAILED DESCRIPTION OF THE INVENTION

[0015] Hereinafter, embodiments of the present disclosure will be described with reference to the drawings. In the description of the drawings, the same elements are given the same reference numerals, and duplicate explanations will be omitted. The drawings may be partially simplified or exaggerated to facilitate understanding, and the dimensional ratios, angles, etc. are not limited to those shown in the drawings.

[0016] FIG. 1 is a side view showing a vehicle 1 equipped with an object detection device 10 according to one embodiment. The vehicle 1 is a large vehicle such as a truck, freight vehicle, or bus, and is typically an autonomous vehicle that travels autonomously without being operated by a driver. In the following, an example will be described in which the vehicle 1 is an autonomous truck. In the following description, the forward and backward directions of the vehicle 1 are referred to as the fore-and-aft direction of the vehicle 1, and the left-to-right direction when the vehicle 1 is viewed from behind is referred to as the vehicle width direction. Furthermore, the direction perpendicular to the vehicle width direction and the fore-and-aft direction is referred to as the up-and-down direction.

[0017] The vehicle 1 is equipped with a laser sensor 2 and an object detection device 10. As shown in FIG. 1, the laser sensor 2 is mounted on the vehicle 1, emits laser light toward an irradiation area R around the vehicle 1, receives reflected laser light, and detects the distance to objects present around the vehicle 1. Objects to be detected include obstacles such as pedestrians, bicycles, other vehicles, and fixed structures (buildings, tunnels, overpasses, signs, plants, etc.). In the embodiment shown in FIG. 1, the laser sensor 2 emits laser light ahead of the vehicle 1, but the direction of laser light emission is not limited to the forward direction. For example, the laser sensor 2 may emit laser light in all directions within a horizontal plane.

[0018] For example, a LiDAR (Laser Imaging Detection and Ranging) is used as the laser sensor 2. The LiDAR scans a laser in the vertical and horizontal directions and outputs position information of an object or environment at each measurement position in three-dimensional coordinates as point cloud data. The point cloud data is a collection of measurement points including measurement results. In other words, each measurement point of the point cloud data includes three-dimensional position information of the object.

[0019] The point cloud data output from the LiDAR includes a point cloud corresponding to the ground surface G. The point cloud corresponding to the ground surface G is a point cloud that indicates the position of the ground surface G. The object detection device 10 removes the point cloud corresponding to the ground surface G from the point cloud data, and analyzes the remaining point cloud to detect an object within the illumination region R.

[0020] The object detection device 10 is mounted on a vehicle 1. The object detection device 10 is a computer including a processor, a storage device, a communication device, and the like. The object detection device 10 realizes various functions described below, for example, by loading a program stored in the storage device and executing the loaded program on the processor. Note that the object detection device 10 does not necessarily have to be configured as a computer that operates according to a program, and some or all of the functions of the object detection device 10 may be implemented in an ASIC (Application Specific Integrated Circuit) that integrates logic circuits.

[0021] 2 is a block diagram showing the functional configuration of an object detection device 10 according to an embodiment. As shown in FIG. 2, the object detection device 10 includes a point cloud data acquisition unit 11 and an object detection unit 12. The point cloud data acquisition unit 11 acquires point cloud data of an object within an irradiation area R generated by a laser sensor 2.

[0022] The object detection unit 12 detects an object based on the acquired point cloud data. As shown in Fig. 2, the object detection unit 12 includes an area setting unit 13, a filtering unit 14, and a clustering unit 15. The area setting unit 13 sets a first monitoring area r1 and a second monitoring area r2 within the irradiation area R. The first monitoring area r1 and the second monitoring area r2 are areas for detecting an object.

[0023] Fig. 3 is a diagram schematically illustrating the first monitoring area r1 and the second monitoring area r2. For convenience, Fig. 3 illustrates the irradiation area R, the first monitoring area r1, and the second monitoring area r2 as two-dimensional areas, but in reality, the irradiation area R, the first monitoring area r1, and the second monitoring area r2 are three-dimensional areas.

[0024] The first monitoring area r1 is a part of the irradiation area R, and is an area located at a position lower than the reference height Hr in the vertical direction. The reference height Hr is a preset value set according to the installation position and installation angle of the laser sensor 2 on the vehicle 1, and is a height position higher than the height of moving objects (pedestrians, bicycles, automobiles, etc.) when the coordinate system of the point cloud data is converted to the world coordinate system. Therefore, the point cloud representing the ground G and the point cloud representing moving objects on the ground G among the point cloud data are included in the first monitoring area r1.

[0025] The first monitoring area r1 has a width W1 in the lateral direction corresponding to the vehicle width direction, which is larger than the width of the vehicle 1 in the vehicle width direction when the coordinate system of the point cloud data is transformed into the world coordinate system.

[0026] The second monitoring area r2 is a part of the illumination area R and is located above the first monitoring area r1. "Located above the first monitoring area r1" means that the center point of the second monitoring area r2 is located above the center point of the first monitoring area r1. That is, the first monitoring area r1 and the second monitoring area r2 are positioned offset in the vertical direction. For example, the second monitoring area r2 is set at a position higher than the reference height Hr in the vertical direction. Therefore, the point cloud representing the ground G and the point cloud representing a moving object on the ground G among the point cloud data are not included in the second monitoring area r2. The first monitoring area r1 and the second monitoring area r2 may partially overlap.

[0027] The second monitoring area r2 has a width W2 in the lateral direction corresponding to the vehicle width direction. When the coordinate system of the point cloud data is converted to a world coordinate system, the width W2 is larger than the width of the vehicle 1 in the vehicle width direction. Furthermore, the width W2 of the second monitoring area r2 is smaller than the width W1 of the first monitoring area r1. In other words, the number of pixels in the lateral direction of the second monitoring area r2 in the LiDAR image is smaller than the number of pixels in the lateral direction of the first monitoring area r1. A non-monitoring area r3, in which no object detection is performed, is set outside the second monitoring area r2 in the lateral direction X.

[0028] The filtering unit 14 executes ground filtering processing to extract a point cloud representing the ground G from a point cloud (first point cloud) included in the first monitoring area r1 among the point cloud data. Ground filtering processing is a technique for separating and removing a point cloud of the ground and a point cloud of non-ground from point cloud data generated by LiDAR. Known algorithms for ground filtering processing include a scan ground filter, a RANSAC ground filter, and a ray ground filter.

[0029] For example, in the scan ground filter, the point cloud included in the first monitoring area r1 is grouped horizontally and sorted by distance from the LiDAR. Next, the lowest measurement point among the sorted point clouds is selected as a candidate for ground G, and measurement points nearby the candidate point for ground G are searched for, and measurement points within a threshold range are classified as measurement points representing ground G. By repeatedly performing the above process, a set of point clouds classified as ground G is extracted as a point cloud representing ground G. The filtering unit 14 outputs a point cloud obtained by removing the point cloud representing ground G from the point cloud included in the first monitoring area r1 to the clustering unit 15.

[0030] On the other hand, the filtering unit 14 does not perform ground filtering on the point cloud included in the second monitoring area r2 of the point cloud data because the second monitoring area r2, which is set at a position higher than the reference height Hr, does not include a point cloud representing the ground G.

[0031] The clustering unit 15 detects an object by performing clustering processing on the point cloud included in the first monitoring area r1 and the point cloud included in the second monitoring area r2. For example, the clustering unit 15 classifies a group of spatially adjacent point clouds into one cluster (group) and recognizes each cluster as an object. Examples of clustering processing methods that are used include the k-means method, DBSCAN (Density-Based Spatial Clustering of Applications with Noise) method, mean shift method, and hierarchical clustering. As described above, the clustering unit 15 performs clustering processing on the point cloud after ground filtering in the first monitoring area r1, and performs clustering processing on the point cloud before ground filtering in the second monitoring area r2.

[0032] The tracking unit 16 tracks the object detected by the clustering unit 15. For example, the tracking unit 16 inputs the position of the object detected by the clustering unit 15 into a Kalman filter and predicts the future position and velocity of the object.

[0033] Information indicating the position and speed of the object predicted by the tracking unit 16 is used for automatic driving of the vehicle 1. For example, a target speed and a target steering angle of the vehicle 1 are determined based on the estimated position and estimated speed of the object, and various actuators of the vehicle 1 are controlled so that the determined target speed and target steering angle are achieved.

[0034] Next, an object detection method according to one embodiment will be described. This object detection method is executed by the above-described object detection device 10. Fig. 4 is a flowchart showing the object detection method according to one embodiment.

[0035] As shown in FIG. 4, in this object detection method, the point cloud data acquisition unit 11 acquires point cloud data of an object in the irradiation region R generated by the laser sensor 2 (step ST1).

[0036] Next, the region setting unit 13 sets a first monitoring region r1 and a second monitoring region r2 within the irradiation region R (step ST2). At this time, the second monitoring region r2 is set at a higher position (above) than the first monitoring region r1.

[0037] Next, the filtering unit 14 performs ground filtering on the point cloud of the first monitoring area r1 (step ST3). As a result, the point cloud representing the ground surface G is removed from the point cloud of the first monitoring area r1. On the other hand, the filtering unit 14 does not perform ground filtering on the point cloud of the second monitoring area r2.

[0038] Next, the clustering unit 15 performs a clustering process on the point cloud of the first monitoring area r1 from which the point cloud representing the ground G has been removed (step ST4). As a result, objects within the first monitoring area r1 are detected. The objects detected within the first monitoring area r1 may be moving objects such as pedestrians, bicycles, and vehicles.

[0039] Next, the clustering unit 15 performs clustering processing on the point cloud of the second monitoring area r2 (step ST5). As a result, objects within the second monitoring area r2 are detected. The objects detected within the second monitoring area r2 are likely to be stationary objects such as buildings and plants. The clustering processing of steps ST4 and ST5 may be performed simultaneously.

[0040] Next, the tracking unit 16 tracks the detected object (step ST6). The position and speed of the object acquired by tracking the object are used for automatic driving.

[0041] As described above, in the object detection device 10, a first monitoring area r1 and a second monitoring area r2 are set within the illumination area R, and ground filtering is performed on the point cloud included in the first monitoring area r1, but not on the point cloud included in the second monitoring area r2. That is, in this object detection device 10, the area in which ground filtering is performed is limited to the first monitoring area r1, thereby reducing the processing load on the object detection device 10. Furthermore, because the second monitoring area is located above the first monitoring area, it is highly likely that the second monitoring area r2 does not include a point cloud representing the ground G. Therefore, even when the area in which ground filtering is performed is limited, it is possible to appropriately remove the point cloud representing the ground G from the point cloud data.

[0042] Furthermore, the width W1 of the first monitoring area r1 in the lateral direction X is set larger than the width W2 of the second monitoring area r2 in the lateral direction X. That is, since the first monitoring area r1 has a wider object detection range in the vehicle width direction, it is possible to predict in advance whether a moving object will jump out from the direction of movement of the moving object detected in the first monitoring area r1. On the other hand, since objects detected in the second monitoring area r2 are likely to be stationary objects such as structures, there is less need to predict whether an object detected in the second monitoring area r2 will jump out. Therefore, by relatively narrowing the width W2 of the second monitoring area r2, it is possible to appropriately detect stationary objects that may come into contact with the vehicle while reducing the processing load.

[0043] The object detection device 10 and object detection method according to various embodiments have been described above, but the invention is not limited to the above-described embodiments and various modifications can be made without departing from the spirit of the invention.

[0044] Although the vehicle 1 has been described as being an autonomous vehicle, the vehicle 1 does not have to be an autonomous vehicle. In this case, information relating to the position and speed of the object detected by the object detection device 10 can be used for driving assistance (such as a function for following a preceding vehicle) for the vehicle 1. The various embodiments described above can be combined to the extent that no contradictions are present.

[0045] The present disclosure includes the following contents.

[0046] [1] An object detection device that irradiates a laser beam toward an irradiation area around a vehicle and detects an object present in the irradiation area using a laser sensor that receives reflected light of the laser beam, a point cloud data acquisition unit that acquires point cloud data of an object in the irradiation area generated by the laser sensor; an object detection unit that detects an object present in the illumination area based on the point cloud data; Equipped with The object detection unit an area setting unit that sets a first monitoring area and a second monitoring area located above the first monitoring area within the irradiation area; a filtering unit that performs ground filtering processing to remove points corresponding to the ground from a first point cloud included in the first monitoring area from the point cloud data; a clustering unit that performs a clustering process on the point cloud after the ground filtering process to detect an object in the first monitoring area, and that performs a clustering process on a second point cloud included in the second monitoring area among the point cloud data without performing the ground filtering process on the second point cloud to detect an object in the second monitoring area; An object detection device comprising:

[0047] [2] The object detection device described in [1], wherein the area setting unit sets the first monitoring area and the second monitoring area so that the width of the second monitoring area in the horizontal direction is smaller than the width of the first monitoring area in the horizontal direction.

[0048] [3] The object detection device described in [1] or [2], wherein the area setting unit sets the first monitoring area and the second monitoring area so that the first monitoring area and the second monitoring area partially overlap.

[0049] [4] The object detection device according to any one of [1] to [3], wherein the laser sensor is a LiDAR.

[0050] [5] An object detection method for detecting an object in an irradiation area around a vehicle by irradiating a laser beam onto the irradiation area and detecting an object present in the irradiation area using a laser sensor that receives reflected light of the laser beam, acquiring point cloud data of an object within the illumination area generated by the laser sensor; detecting an object present in the illumination area based on the point cloud data; Including, The step of detecting an object includes: setting a first monitoring area and a second monitoring area located above the first monitoring area within the irradiation area; performing ground filtering processing to remove points corresponding to the ground from a first point cloud included in the first monitoring area from the point cloud data; performing a clustering process on the point cloud after the ground filtering process to detect an object in the first monitoring area, and performing a clustering process on a second point cloud included in the second monitoring area of ​​the point cloud data without performing the ground filtering process on the second point cloud to detect an object in the second monitoring area; 1. An object detection method comprising: [Explanation of symbols]

[0051] 1...vehicle, 2...laser sensor, 10...object detection device, 11...point cloud data acquisition unit, 12...object detection unit, 13...area setting unit, 14...filtering unit, 15...clustering unit, G...ground, R...illumination area, r1...first monitoring area, r2...second monitoring area.

Claims

1. An object detection device that irradiates a laser beam toward an irradiation area around a vehicle and detects an object present in the irradiation area using a laser sensor that receives reflected light of the laser beam, a point cloud data acquisition unit that acquires point cloud data of an object in the irradiation area generated by the laser sensor; an object detection unit that detects an object present in the illumination area based on the point cloud data; Equipped with The object detection unit an area setting unit that sets a first monitoring area and a second monitoring area located above the first monitoring area within the irradiation area; a filtering unit that performs ground filtering processing to remove points corresponding to the ground from a first point cloud included in the first monitoring area from the point cloud data; a clustering unit that performs a clustering process on the point cloud after the ground filtering process to detect an object in the first monitoring area, and that performs a clustering process on a second point cloud included in the second monitoring area of ​​the point cloud data without performing the ground filtering process on the second point cloud to detect an object in the second monitoring area; An object detection device comprising:

2. 2. The object detection device according to claim 1, wherein the region setting unit sets the first monitoring region and the second monitoring region so that a width of the second monitoring region in the horizontal direction is smaller than a width of the first monitoring region in the horizontal direction.

3. The object detection device according to claim 1 , wherein the region setting unit sets the first monitoring region and the second monitoring region so that the first monitoring region and the second monitoring region partially overlap each other.

4. The object detection device according to claim 1 , wherein the laser sensor is a LiDAR.

5. 1. An object detection method for detecting an object in an irradiation area around a vehicle by irradiating a laser beam onto the irradiation area and detecting an object present in the irradiation area using a laser sensor that receives reflected light of the laser beam, acquiring point cloud data of an object within the illumination area generated by the laser sensor; detecting an object present in the illumination area based on the point cloud data; Including, The step of detecting an object includes: setting a first monitoring area within the irradiation area and a second monitoring area positioned above the first monitoring area; performing ground filtering processing to remove points corresponding to the ground from a first point cloud included in the first monitoring area from the point cloud data; performing a clustering process on the point cloud after the ground filtering process to detect an object in the first monitoring area, and performing a clustering process on a second point cloud included in the second monitoring area of ​​the point cloud data without performing the ground filtering process on the second point cloud to detect an object in the second monitoring area; 1. An object detection method comprising:

Images