WO2020206639A1 - 目标物体的拟合方法、点云传感器和移动平台 - Google Patents
目标物体的拟合方法、点云传感器和移动平台 Download PDFInfo
- Publication number
- WO2020206639A1 WO2020206639A1 PCT/CN2019/082119 CN2019082119W WO2020206639A1 WO 2020206639 A1 WO2020206639 A1 WO 2020206639A1 CN 2019082119 W CN2019082119 W CN 2019082119W WO 2020206639 A1 WO2020206639 A1 WO 2020206639A1
- Authority
- WO
- WIPO (PCT)
- Prior art keywords
- angle
- fitting
- point cloud
- visible
- area
- Prior art date
Links
- 238000000034 method Methods 0.000 title claims abstract description 68
- 238000004364 calculation method Methods 0.000 claims description 5
- 238000004891 communication Methods 0.000 claims description 5
- 238000010586 diagram Methods 0.000 description 15
- 238000011478 gradient descent method Methods 0.000 description 5
- 238000004590 computer program Methods 0.000 description 4
- 230000002776 aggregation Effects 0.000 description 2
- 238000004220 aggregation Methods 0.000 description 2
- 230000000694 effects Effects 0.000 description 2
- 230000001788 irregular Effects 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 230000003287 optical effect Effects 0.000 description 1
- 230000008447 perception Effects 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/20—Image preprocessing
- G06V10/255—Detecting or recognising potential candidate objects based on visual cues, e.g. shapes
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T3/00—Geometric image transformations in the plane of the image
- G06T3/06—Topological mapping of higher dimensional structures onto lower dimensional surfaces
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/20—Analysis of motion
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V20/00—Scenes; Scene-specific elements
- G06V20/50—Context or environment of the image
- G06V20/56—Context or environment of the image exterior to a vehicle by using sensors mounted on the vehicle
- G06V20/58—Recognition of moving objects or obstacles, e.g. vehicles or pedestrians; Recognition of traffic objects, e.g. traffic signs, traffic lights or roads
- G06V20/584—Recognition of moving objects or obstacles, e.g. vehicles or pedestrians; Recognition of traffic objects, e.g. traffic signs, traffic lights or roads of vehicle lights or traffic lights
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/10—Image acquisition modality
- G06T2207/10028—Range image; Depth image; 3D point clouds
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/30—Subject of image; Context of image processing
- G06T2207/30248—Vehicle exterior or interior
- G06T2207/30252—Vehicle exterior; Vicinity of vehicle
Definitions
- This application relates to the field of automatic driving, and in particular to a method for fitting a target object, a point cloud sensor and a mobile platform.
- vehicle point cloud is a set of discrete points. In actual processing, it is often necessary to fit the point cloud determined as the target vehicle to obtain the fitted vehicle to determine the orientation and size of the vehicle according to the fitted vehicle. Complete the tracking task.
- the embodiments of the present application provide a method for fitting a target object, a point cloud sensor and a mobile platform, which improve the accuracy of fitting the target object.
- an embodiment of the present application provides a method for fitting a target object, including: generating a two-dimensional point cloud in a top view according to a point cloud of the target object collected by a point cloud sensor, wherein the point cloud sensor is used for carrying In the mobile platform; determine the visible area of the two-dimensional point cloud toward the mobile platform; determine the fitting feature of the target object according to the visible area, and perform the target object according to the fitting feature Fitting.
- determining the fitting feature of the target object according to the visible area includes: determining according to the visible area and the two-dimensional point cloud The fitting feature.
- the fitting feature is the fitting angle of the target object, and the determining the fitting angle is based on the visible area and the two-dimensional point cloud.
- the fitting feature includes: determining the fitting angle according to the visible area, the two-dimensional point cloud, and a preset angle range.
- the determining the fitting angle according to the visible area, the two-dimensional point cloud, and a preset angle range includes: according to the The visible area, the two-dimensional point cloud, and the preset angle range determine the fitting angle with the least corresponding visible edge cost.
- the fitting is determined according to the visible area, the two-dimensional point cloud, and a preset angle range.
- the angle includes: selecting at least one reference angle from the preset angle range; according to the at least one reference angle, the visible area, and the two-dimensional point cloud, determining the corresponding fitting with the least visible edge cost angle.
- Determining the corresponding fitting angle with the smallest visible edge cost includes: for any one of a plurality of reference angles, according to the visible area and the two-dimensional point cloud, in the first reference angle Determine the first preselected angle with the smallest visible edge cost within the first neighborhood range, where the first neighborhood range includes the first reference angle; determine the corresponding preselected angle with the least visible edge cost among the preselected angles as said Fitting angle.
- the visible edge cost is the smallest within the first neighborhood of the first reference angle
- the first preselected angle includes: obtaining a first visible edge cost corresponding to the first reference angle according to the visible area and the two-dimensional point cloud, and updating the first reference angle to the first neighbor
- the second visible edge cost corresponding to the first angle is obtained according to the visible area and the two-dimensional point cloud
- the second visible edge cost and the first visible edge The relationship between the costs, update the first angle to the second angle in the first neighborhood; repeat the execution of obtaining the visible edge costs corresponding to the angles in the first neighborhood, according to two adjacent
- the obtained relationship of the visible edge cost is an operation of updating the angle within the first neighborhood range until the first preselected angle with the smallest visible edge cost is obtained.
- Determining the corresponding fitting angle with the smallest visible edge cost includes: determining the corresponding target reference angle with the smallest visible edge cost among multiple reference angles according to the visible area and the two-dimensional point cloud; For the visible area and the two-dimensional point cloud, determine the fitting angle with the smallest visible edge cost within a second neighborhood range of the target reference angle, and the second neighborhood range includes the target reference angle .
- the fitting angle includes: acquiring a plurality of sub-reference angles from within a second neighborhood of the target reference angle; determining the visible edge cost corresponding to each sub-reference angle according to the visible area and the two-dimensional point cloud The smallest sub-reference angle is the fitting angle.
- obtaining the first visible edge cost corresponding to the first reference angle includes: The first outer envelope rectangle of the two-dimensional point cloud in the first direction, the visible area and the first boundary and the second boundary corresponding to the visible area, to obtain the first visible edge cost; wherein, The first direction is a direction corresponding to the rotation of the first reference angle from the coordinate origin of the point cloud sensor coordinate system, and the coordinate origin of the point cloud sensor coordinate system is the position of the point cloud sensor.
- the visible area corresponds to the visible area
- Obtaining the first visible edge cost includes: obtaining the first area of the first area defined by the visible area, the first border and the second border; obtaining the The first outer envelope rectangle faces the visible side of the mobile platform, the second area of the second area defined by the first boundary and the second boundary; obtains the side of the first envelope rectangle outside the target area, A third area of a third area defined by the first boundary and the second boundary, the target area is the area between the first boundary and the second boundary, and the target area includes the two-dimensional point cloud ; According to the first area, the second area and the third area, determine the first visible edge cost.
- the first visible edge cost includes: obtaining the first area, the second area, and the first preset weight, and obtaining the first preselected visible edge cost; according to the third area and the second preset weight, obtaining the first Two preselected visible edge costs; determining the first visible edge cost according to the first preselected visible edge cost and the second preselected visible edge cost.
- the angle included in the preset angle range is an angle between a first preset angle and a second preset angle, and the first preset angle The angle minus the second preset angle is equal to 90°, and the preset angle range includes the first preset angle and/or the second preset angle.
- the fitting feature includes at least one of the following: a fitting angle of the target object, a fitting direction of the target object, and The fitted bounding box of the target object, and the fitted visible edge of the target object;
- the fitting direction is the direction corresponding to the rotation of the fitting angle from the coordinate origin of the point cloud sensor coordinate system
- the coordinate origin of the point cloud sensor coordinate system is the position of the point cloud sensor
- the fitting the target object according to the fitting feature includes:
- the fitting feature includes at least one of the following: the fitting angle of the target object, and the fitting direction of the target object , Fitting visible edges of the target object, fitting the target object according to the fitting feature and the fitting height, including:
- obtaining the fitting height of the target object according to the point cloud of the target object includes:
- the fitting height is obtained.
- an embodiment of the present application provides a point cloud sensor, which is used to be mounted on a mobile platform, and the point cloud sensor includes:
- the processor is in communication connection with the collector, and is configured to execute the method described in the first aspect and any possible implementation manner of the first aspect. It is understandable that, in the second aspect, "generate a two-dimensional point cloud in the top view according to the point cloud of the target object collected by the point cloud sensor" is "the point cloud of the target object collected by the collector of the point cloud sensor, Generate a two-dimensional point cloud under the top view”.
- embodiments of the present application provide a mobile platform, including: a point cloud sensor for collecting the point cloud of the target object; wherein the point cloud sensor is mounted on the mobile platform; a processor, and The point cloud sensor communication connection is used to execute the method described in the first aspect and any possible implementation manner of the first aspect.
- an embodiment of the present application provides a mobile platform, characterized in that the point cloud sensor described in the second aspect is mounted on the mobile platform.
- an embodiment of the present application provides a computer-readable storage medium, including a program or instruction.
- the program or instruction runs on a computer, the first aspect and any possible implementation of the first aspect The method is executed.
- an embodiment of the present invention provides a computer-readable storage medium, the computer-readable storage medium stores a computer program, the computer program includes at least one piece of code, the at least one piece of code can be executed by a computer to control the computer to execute
- the computer program includes at least one piece of code, the at least one piece of code can be executed by a computer to control the computer to execute
- an embodiment of the present invention provides a computer program, when the computer program is executed by a computer, it is used to execute the foregoing first aspect and the method described in any possible implementation manner of the first aspect.
- the program may be stored in whole or in part on a storage medium that is packaged with the processor, and may also be stored in part or all in a storage medium that is not packaged with the processor.
- the storage medium is, for example, a memory.
- the target object Since the target object is close to the point cloud sensor or one or two surfaces of the mobile platform equipped with the point cloud sensor can be observed by the point cloud sensor, it is away from the point cloud sensor or two or three surfaces of the mobile platform equipped with the point cloud sensor It cannot be observed by the point cloud sensor. Therefore, the point cloud sensor based on the target object in this application can be fitted to the target object, that is, according to the visible area of the target object’s two-dimensional point cloud towards the mobile platform, the target Object fitting can improve the accuracy of fitting the target object.
- Figure 1 is a schematic diagram of an application scenario provided by an embodiment of the application
- FIG. 2 is a first flowchart of a method for fitting a target object provided by an embodiment of the application
- FIG. 3 is a schematic diagram of the visible area facing the mobile platform in the two-dimensional point cloud provided by an embodiment of the application;
- FIG. 4 is a schematic diagram 1 of fitting visible edges of a target object provided by the implementation of this application.
- 5 is a second schematic diagram of fitting visible edges of the target object provided by the implementation of this application.
- FIG. 6 is the third schematic diagram of fitting visible edges of the target object provided by the implementation of this application.
- FIG. 7 is a second flowchart of a method for fitting a target object provided by an embodiment of the application.
- FIG. 8 is a schematic diagram of obtaining visible edge cost provided by an embodiment of the application.
- FIG. 9 is a schematic structural diagram of a point cloud sensor provided by an embodiment of the application.
- Fig. 10 is a schematic structural diagram of a mobile platform provided by an embodiment of the application.
- FIG. 1 is a schematic diagram of an application scenario provided by an embodiment of the application.
- a cloud sensor 12 is mounted on a mobile platform 11.
- the point cloud sensor 12 is used to collect point cloud data of the target object, and the point cloud data of the target object is used to fit the target object.
- FIG. 2 is a first flowchart of a method for fitting a target object provided by an embodiment of the application. Referring to Fig. 2, the method of this embodiment includes:
- Step S101 Generate a two-dimensional point cloud in the top view according to the point cloud of the target object collected by the point cloud sensor, where the point cloud sensor is used to be mounted on the mobile platform.
- the point cloud of the target object collected by the point cloud sensor is a three-dimensional point cloud
- the three-dimensional point cloud can be projected on a plane in the vertical direction to obtain a two-dimensional point cloud, that is, the two-dimensional point cloud under the top view is generated according to the point cloud of the target object Point cloud.
- the target object is a vehicle driving on the road
- the collected three-dimensional point cloud of the vehicle can be projected on the plane of the road along the vertical direction, so as to obtain the two-dimensional point cloud in the top view.
- the point cloud sensor in this embodiment may be a TOF sensor or a lidar.
- the mobile platform in this embodiment may be a vehicle, and the target object may be other vehicles driving on the road.
- the vehicle is equipped with a lidar, which may be one or more lidars, and may be a rotating lidar or a solid-state lidar. The lidar is used to obtain three-dimensional point cloud information around the vehicle.
- Step S102 Determine the visible area facing the mobile platform in the two-dimensional point cloud.
- the present embodiment performs fitting based on the face of the target object that can be observed by the point cloud sensor of the target object, that is, to determine the two-dimensional
- the visible area facing the mobile platform in the point cloud is used to fit the target object according to the visible area facing the mobile platform in the two-dimensional point cloud.
- the visible area of the two-dimensional point cloud of the target object facing the mobile platform is the area of the two-dimensional point cloud of the target object in the sensing range and the field of view of the point cloud sensor.
- FIG. 3 is a schematic diagram of a visible area facing a mobile platform in a two-dimensional point cloud provided by an embodiment of the application.
- Fig. 3 which illustrates some points in the two-dimensional point cloud.
- the coordinate origin O shown in Fig. 3 is the position of the point cloud sensor
- the irregular figure 21 in Fig. 3 is the contour or convex hull of the two-dimensional point cloud of the target object, and the contour or convex hull of the two-dimensional point cloud is added
- the thick part 211 is the visible area facing the mobile platform in the two-dimensional point cloud
- OA and OB are the boundaries corresponding to the visible area facing the mobile platform in the two-dimensional point cloud.
- Step S103 Determine the fitting feature of the target object according to the visible area facing the mobile platform in the two-dimensional point cloud of the target object, and fit the target object according to the fitting feature.
- the fitting feature of the target object can be determined according to the visible area, and the target object can be fitted according to the fitting feature.
- fitting the target object according to the fitting feature includes:
- the fitted maximum height and the fitted minimum height of the point cloud of the target object are obtained, and the fitted height of the target object is obtained according to the fitted maximum height and the fitted minimum height.
- the difference between the maximum fitting height and the minimum fitting height is the fitting height of the target object.
- the maximum fitting height can be the height of the point with the highest height in the point cloud of the target object, or the maximum height obtained according to hierarchical aggregation, such as fitting according to the height of all points in the point cloud with a certain thickness
- the point cloud of a certain thickness includes the point with the largest height in the point cloud of the target object.
- the minimum fitting height can be the height of the point with the smallest height in the point cloud of the target object, or the minimum height obtained according to hierarchical aggregation, such as fitting according to the height of all points in a point cloud with a certain thickness. It can be understood that the point cloud of a certain thickness includes the point with the smallest height in the point cloud of the target object.
- the fitting feature of the target object may include at least one of the following: fitting angle of the target object, fitting direction of the target object, fitting bounding box of the target object, and fitting visible edge of the target object.
- the fitting feature of the target object may also be other features, which is not limited in this embodiment.
- fitting feature of the target object is the fitting angle of the target object
- fitting the target object according to the fitting feature of the target object and the fitting height of the target object including:
- the direction corresponding to the rotation fitting angle from the coordinate origin of the point cloud sensor coordinate system is the fitting direction of the target object, and the coordinate origin of the point cloud sensor coordinate system is the position of the point cloud sensor.
- the preset angle range and the preselected angle range may be the same or different; the angle included in the preset angle range is the angle between the first preset angle and the second preset angle, and the second preset angle minus the first preset angle Set the angle equal to 90°, the preset angle range includes the first preset angle and/or the second preset angle; the preselected angle range includes the angle between the third preset angle and the fourth preset angle, and the fourth The preset angle minus the third preset angle is equal to 90°, and the preselected angle range includes the third preset angle and/or the fourth preset angle.
- the pre-selected angle range is obtained according to the preset angle range.
- the fitting angle of the target object is the angle corresponding to the minimum cost of the visible edge.
- each angle corresponds to a visible edge cost.
- the visible edge cost is the degree of fit between the outer envelope rectangle of the two-dimensional point cloud of the target object in the direction corresponding to the angle and the two-dimensional point cloud, and the overflow of the outer envelope rectangle The parameter of the overflow degree of the target area; where the direction corresponding to the angle is the direction corresponding to the angle after rotating the angle from the coordinate origin of the point cloud sensor coordinate system.
- the target area is the area between the first boundary and the second boundary corresponding to the visible area facing the mobile platform in the two-dimensional point cloud of the target object, and the target area includes the two-dimensional point cloud of the target object; further, the The first boundary and the second boundary can surround the two-dimensional point cloud, and both the first boundary and the second boundary have intersection points with the convex hull of the two-dimensional point cloud.
- OA in Figure 3 is the first boundary corresponding to the visible area facing the mobile platform in the two-dimensional point cloud of the target object
- OB is the first boundary corresponding to the visible area facing the mobile platform in the two-dimensional point cloud of the target object. Two borders.
- the outer envelope rectangle of the two-dimensional point cloud of the target object in the direction corresponding to the angle satisfies the following conditions: the two sides of the outer envelope rectangle are in the same direction as the direction corresponding to the angle, and the other The direction of the two sides is perpendicular to the direction corresponding to the angle.
- the outer envelope rectangle of the two-dimensional point cloud of the target object in the fitting direction is obtained, and the outer envelope rectangle is the fitting bounding box of the target object.
- the outer envelope rectangle of the two-dimensional point cloud of the target object in the fitting direction satisfies the following conditions: the two sides of the outer envelope rectangle are in the same direction as the fitting direction of the target object, and the outer envelope The direction of the other two sides of the rectangle is perpendicular to the fitting direction of the target object.
- the fitting direction of the target object in the process of obtaining the fitted bounding box of the target object, first obtain the fitted visible edge of the target object according to the fitting direction, and then use the fitted visible edge of the target object as the edge to obtain the target
- the outer envelope rectangle of the two-dimensional point cloud of the object in the fitting direction in the process of obtaining the fitted bounding box of the target object, first obtain the fitted visible edge of the target object according to the fitting direction, and then use the fitted visible edge of the target object as the edge to obtain the target The outer envelope rectangle of the two-dimensional point cloud of the object in the fitting direction.
- the fitted visible edge of the target object can be referred to as the L-shaped feature or the I-shaped feature of the target object. It is understandable that the fitted visible edge of the target object satisfies the following conditions: (1) The two-dimensional point cloud of the target object is located on the same side of the fitted visible edge of the target object, and the convex hull of the two-dimensional point cloud of the target object is The fitted visible edges of the target object have intersections; (2) In the case of two fitted visible edges of the target object, one of the two fitted visible edges is located in the same direction as the fitting direction of the target object , The direction of the other edge is perpendicular to the fitting direction of the target object; in the case that the fitted visible edge of the target object has one, the fitted visible edge is located in the same direction as the fitting direction of the target object.
- FIG. 4 is the first schematic diagram of the fitted visible edge of the target object provided by the implementation of this application
- FIG. 5 is the second schematic diagram of the fitted visible edge of the target object provided by the implementation of this application
- FIG. 6 is the simulated view of the target object provided by the implementation of this application Schematic three of the visible edges.
- 41 and 42 are the fitted visible edges of the target object, and the fitted visible edges at this time are the L-shaped features of the target object.
- 51 is the fitted visible edge of the target object, and the fitted visible edge at this time is the I-type feature of the target object.
- 61 is the fitted visible edge of the target object, and the fitted visible edge at this time is the I-type feature of the target object.
- the fitting feature of the target object is the fitting direction of the target object
- the fitting feature of the target object is acquired according to the visible area of the two-dimensional point cloud of the target object toward the mobile platform, including:
- fitting the target object according to the fitting feature of the target object and the fitting height of the target object including:
- the fitting feature of the target object is the fitting bounding box of the target object
- the fitting feature of the target object is determined according to the visible area of the two-dimensional point cloud of the target object toward the mobile platform, including:
- fitting the target object according to the fitting feature of the target object and the fitting height of the target object including: according to the fitting bounding box of the target object and the fitting height of the target object, The object is fitted.
- the fitting feature of the target object is the fitting visible edge of the target object
- the fitting feature of the target object is obtained according to the visible area of the two-dimensional point cloud of the target object toward the mobile platform, including:
- fitting the target object according to the fitting feature of the target object and the fitting height of the target object including:
- the point cloud sensor based on the target object can perform fitting to the target object, that is, according to the visible area facing the mobile platform in the two-dimensional point cloud of the target object, Fitting the target object can improve the accuracy of fitting the target object.
- FIG. 7 is the second flowchart of the target object fitting method provided by the embodiment of the application. Referring to FIG. 7, the method of this embodiment includes:
- Step S201 Select at least one reference angle from a preset angle range.
- the meaning of the preset angle range in this embodiment is the same as the meaning of the preset angle range in the previous embodiment.
- At least one reference angle refers to one or more reference angles.
- the reference angle can be any angle within the first preset reference range.
- the preset angle range is [0°, 90°]
- one reference angle may be 45°.
- one reference angle can be selected at the same interval within the preset angle range to obtain multiple reference angles.
- the multiple reference angles may include: 0°, 10°, 20°, 30°, 40°, 50°, 60°, 70°, 80°, 90°
- multiple reference angles can include: 0°, 10°, 20°, 30°, 40°, 50°, 60°, 70 °, 80°.
- Step S202 According to at least one reference angle, the visible area facing the mobile platform in the two-dimensional point cloud of the target object, and the two-dimensional point cloud, determine the corresponding fitting angle with the least visible edge cost.
- the following takes multiple reference angles as an example to illustrate the method of “determining the corresponding fitting angle with the least visible edge cost based on the multiple reference angles and the visible area”.
- the first preselected angle includes the first reference angle in the first neighborhood.
- each reference angle has a neighborhood range, and therefore has multiple neighborhood ranges, and the union of the multiple neighborhood ranges is the preselected angle range in the previous embodiment.
- the preset angle range is [P 1 °,P 2 °) or [P 1 °,P 2 °] or (P 1 °, P 2 °], and a reference is selected at intervals of ⁇ ° within the preset angle range
- the first reference angle is N°
- the first neighborhood range is (N°- ⁇ °/2, N°+ ⁇ °/2] or [N°- ⁇ ° /2, N°+ ⁇ °/2)
- the preset angle range and the preselected angle range are not the same; in another way, if the first reference angle is N°, if N is not equal to P 1 and neither When equal to P 2 , the first neighborhood range is (N°- ⁇ °/2, N°+ ⁇ °/2] or [N°- ⁇ °/2, N°+ ⁇ °/2), where N is equal to
- the first neighborhood range is [N°, N°+ ⁇ °/2] or (N°, N°+ ⁇ °/2] or [N°, N°+ ⁇ °/
- the gradient descent method can be used to determine the first preselected angle with the least visible edge cost within the first neighborhood of the first reference angle, specifically:
- the method for obtaining the first visible edge cost is as follows: According to the first outer envelope rectangle of the two-dimensional point cloud in the first direction, the visible area and the first boundary and the second boundary corresponding to the visible area, obtain The first visible edge cost corresponding to the first reference angle; where the first direction is the direction corresponding to the rotation of the first reference angle from the coordinate origin of the point cloud sensor coordinate system.
- the first boundary and the second boundary corresponding to the visible area are the same as those described in the previous embodiment.
- FIG. 8 is a schematic diagram of obtaining the visible edge cost provided by an embodiment of the application.
- the following is a description of "according to the first outer envelope rectangle of the two-dimensional point cloud in the first direction, the visible area and the first visible area corresponding to the visible area in conjunction with FIG. A boundary and a second boundary, the process of obtaining the first visible edge cost corresponding to the first reference angle is described.
- "According to the first outer envelope rectangle of the two-dimensional point cloud in the first direction, the visible area and the first and second boundaries corresponding to the visible area, obtain the first visible edge corresponding to the first reference angle Cost" including:
- OA in FIG. 8 is the first boundary
- OB is the second boundary
- the first area is the area 81 in FIG. 8.
- the first direction is the direction indicated by the arrow in FIG. 8
- the second area is the area 82 in FIG. 8.
- the target area includes a two-dimensional point cloud of the target object.
- the third area is the area 83 in FIG. 8, that is, the area filled with solid lines.
- determining the visible edge cost corresponding to the first reference angle according to the first area, the second area, and the third area includes:
- the first preselected visible edge cost is used to characterize the degree of fit between the first outer envelope rectangle and the convex hull of the two-dimensional point cloud.
- the first preselected visible edge cost is equal to the difference between the first area and the second area multiplied by the first preset weight.
- the difference between the first area and the second area is the area of the area filled by the dotted line in the figure.
- the second preselected visible edge cost is used to characterize the overflow degree of the first outer envelope rectangle overflowing the target area.
- the second preselected visible edge cost is equal to the third area multiplied by the second preset weight.
- the visible edge cost is a combination of the fit between the outer envelope rectangle of the two-dimensional point cloud of the target object in the direction corresponding to a certain angle and the two-dimensional point cloud and the overflow degree of the outer envelope rectangle overflowing the target area. parameter.
- the first angle may be an angle greater than the first reference angle, or may be an angle smaller than the first reference angle.
- the method of obtaining the second visible edge cost corresponding to the first angle is the same as the method of obtaining the first visible edge cost corresponding to the first reference angle, and will not be repeated here.
- angle update rules are as follows:
- first angle is greater than the first reference angle, and the first visible edge cost is greater than the second visible edge cost, update the first angle to a second angle smaller than the first angle in the first neighborhood.
- first angle is greater than the first reference angle, and the first visible edge cost is less than the second visible edge cost, update the first angle to a second angle larger than the first angle in the first neighborhood.
- first angle is smaller than the first reference angle, and the first visible edge cost is greater than the second visible edge cost, update the first angle to a second angle larger than the first angle in the first neighborhood.
- first angle is smaller than the first reference angle, and the first visible edge cost is smaller than the second visible edge cost, update the first angle to a second angle smaller than the first angle in the first neighborhood.
- the above j1 to j3 are specific implementations of using the gradient descent method to determine the first preselected angle with the smallest visible edge cost within the first neighborhood of the first reference angle.
- the preselected angles corresponding to the other reference angles are obtained according to the method of obtaining the first preselected angle corresponding to the first reference angle, and finally multiple preselected angles are obtained.
- the fitting angle of the target object with the least visible edge cost is determined according to at least one reference angle, the visible area and the two-dimensional point cloud of the target object ,include:
- m1 according to the visible area and the two-dimensional point cloud of the target object, determine the corresponding target reference angle with the least visible edge cost among multiple reference angles.
- the visible edge cost corresponding to each reference angle is obtained, and the reference angle with the smallest corresponding visible edge cost is determined as the target reference angle.
- obtaining the visible edge cost corresponding to each reference angle refers to the method for obtaining the visible edge cost corresponding to the first reference angle.
- the second neighborhood range is (S 1 °, S 2 °), where S 1 ° is the absolute value of the difference with the target reference angle being the smallest and smaller than the target reference angle S 2 ° is the reference angle with the smallest absolute value of the difference from the target reference angle and greater than the target reference angle.
- the gradient descent method can be used to determine the fitting angle of the target object with the least visible edge cost within the second neighborhood of the target reference angle based on the visible area and the two-dimensional point cloud of the target object .
- the specific implementation of this manner refers to the method of determining the first preselected angle in the gradient descent method within the first neighborhood of the first reference angle, which will not be repeated here.
- multiple sub-reference angles are obtained from the second neighborhood of the target reference angle, and the smallest visible edge cost corresponding to each sub-reference angle is determined according to the visible area and the two-dimensional point cloud of the target object.
- the sub-reference angle is the fitting angle of the target object.
- obtaining the visible edge cost corresponding to each sub-reference angle refers to the obtaining method of obtaining the visible edge cost corresponding to the first reference angle, which will not be repeated here.
- a sub-reference angle may be obtained at every interval of the same angle within the second neighborhood range of the target reference angle.
- This embodiment provides a specific method for obtaining the fitting angle of the target object, and the method of this embodiment can further improve the accuracy of obtaining the fitting angle of the target object.
- FIG. 9 is a schematic structural diagram of a point cloud sensor provided by an embodiment of the application.
- the point cloud sensor provided in this embodiment is used to be mounted on a mobile platform.
- the point cloud sensor includes a collector 91 and a processor 92.
- the collector 91 is used to collect the point cloud of the target object
- the processor 92 is communicatively connected to the collector and is configured to perform the following operations: generate a two-dimensional point cloud in the top view according to the point cloud of the target object collected by the collector; and determine the direction in the two-dimensional point cloud The visible area of the mobile platform; according to the visible area, the fitting feature of the target object is determined, and the target object is fitted according to the fitting feature.
- the processor 92 when configured to perform the operation of determining the fitting feature of the target object according to the visible area, it is specifically configured to: determine according to the visible area and the two-dimensional point cloud The fitting feature.
- the fitting feature is the fitting angle of the target object
- the processor 92 is configured to perform the operation of determining the fitting feature based on the visible area and the two-dimensional point cloud , Specifically used for: determining the fitting angle according to the visible area, the two-dimensional point cloud, and a preset angle range.
- the processor 92 when configured to perform the operation of determining the fitting angle according to the visible area, the two-dimensional point cloud, and a preset angle range, it is specifically configured to: , The two-dimensional point cloud and the preset angle range determine the corresponding fitting angle with the smallest visible edge cost.
- the processor 92 when the processor 92 is configured to perform the operation of determining the corresponding fitting angle with the least visible edge cost according to the visible area, the two-dimensional point cloud, and a preset angle range, specifically use In: selecting at least one reference angle from the preset angle range; and determining the corresponding fitting angle with the least visible edge cost according to the at least one reference angle, the visible area, and the two-dimensional point cloud.
- the processor 92 is configured to determine that the corresponding visible edge cost is the smallest according to the at least one reference angle, the visible area, and the two-dimensional point cloud.
- the fitting angle it is specifically used to: for any one of the multiple reference angles, according to the visible area and the two-dimensional point cloud, in the first reference angle Determine a first preselected angle with the smallest visible edge cost within a neighborhood range, where the first neighborhood range includes the first reference angle; determine the corresponding preselected angle with the least visible edge cost among the preselected angles as the fitting angle.
- the processor 92 is configured to perform determining a first preselected angle with the least visible edge cost within a first neighborhood range of the first reference angle based on the visible area and the two-dimensional point cloud During the operation, it is specifically used to obtain the first visible edge cost corresponding to the first reference angle according to the visible area and the two-dimensional point cloud, and update the first reference angle to the first neighbor According to the first angle within the domain, the second visible edge cost corresponding to the first angle is obtained according to the visible area and the two-dimensional point cloud; according to the second visible edge cost and the first visible edge The relationship between the costs, update the first angle to the second angle in the first neighborhood; repeat the execution of obtaining the visible edge costs corresponding to the angles in the first neighborhood, according to two adjacent The obtained relationship of the visible edge cost is an operation of updating the angle within the first neighborhood range until the first preselected angle with the smallest visible edge cost is obtained.
- the processor 92 is configured to determine that the corresponding visible edge cost is the smallest according to the at least one reference angle, the visible area, and the two-dimensional point cloud.
- the fitting angle it is specifically used to: determine the corresponding target reference angle with the least visible edge cost among multiple reference angles according to the visible area and the two-dimensional point cloud; according to the visible area And the two-dimensional point cloud, determining the corresponding fitting angle with the smallest visible edge cost within a second neighborhood range of the target reference angle, and the second neighborhood range includes the target reference angle.
- the processor 92 is configured to determine the fitting angle with the least visible edge cost within the second neighborhood range of the target reference angle according to the visible area and the two-dimensional point cloud During the operation, it is specifically used to: obtain multiple sub-reference angles from within the second neighborhood of the target reference angle; determine the visible edge cost corresponding to each sub-reference angle according to the visible area and the two-dimensional point cloud The smallest sub-reference angle is the fitting angle.
- the processor 92 when configured to perform the operation of obtaining the first visible edge cost corresponding to the first reference angle according to the visible area and the two-dimensional point cloud, it is specifically configured to: The first outer envelope rectangle of the two-dimensional point cloud in the first direction, the visible area and the first boundary and the second boundary corresponding to the visible area, to obtain the first visible edge cost; wherein, The first direction is a direction corresponding to the rotation of the first reference angle from the coordinate origin of the point cloud sensor coordinate system, and the coordinate origin of the point cloud sensor coordinate system is the position of the point cloud sensor.
- the processor 92 is configured to execute the first outer envelope rectangle in the first direction according to the two-dimensional point cloud, the visible area and the first boundary and the second boundary corresponding to the visible area.
- Boundary when the operation of obtaining the first visible edge cost is specifically used to: obtain the first area of the first area defined by the visible area, the first boundary, and the second boundary; The outer envelope rectangle faces the visible side of the mobile platform, and the second area of the second area defined by the first boundary and the second boundary; obtains the side of the first envelope rectangle outside the target area, the A third area of a third area defined by the first boundary and the second boundary, the target area is the area between the first boundary and the second boundary, and the target area includes the second boundary of the target object.
- One-dimensional point cloud determine the first visible edge cost according to the first area, the second area, and the third area.
- the processor 92 is configured to perform an operation of determining the first visible edge cost corresponding to the first reference angle according to the first area, the second area, and the third area. , It is specifically used to: obtain the first area, the second area, and the first preset weight, obtain the first preselected visible edge cost; according to the third area and the second preset weight, obtain the second preselect Visible edge cost; determining the first visible edge cost according to the first preselected visible edge cost and the second preselected visible edge cost.
- the angle included in the preset angle range is an angle between a first preset angle and a second preset angle, and the first preset angle minus the second preset angle is equal to 90°,
- the preset angle range includes the first preset angle and/or the second preset angle.
- the fitting feature includes at least one of the following: the fitting angle of the target object, the fitting direction of the target object, the fitting bounding box of the target object, and the fitting of the target object.
- the processor 92 when configured to perform the operation of fitting the target object according to the fitting feature, it is specifically configured to: obtain the information of the target object according to the point cloud of the target object Fitting height; fitting the target object according to the fitting feature and the fitting height.
- the processing when the fitting feature includes at least one of the following: the fitting angle of the target object, the fitting direction of the target object, and the fitting visible edge of the target object, the processing When the device 92 is used to perform the operation of fitting the target object according to the fitting feature and the fitting height, it is specifically configured to: obtain the fitting boundary of the target object according to the fitting feature Box; according to the fitting bounding box and fitting height, fitting the target object.
- the processor 92 when configured to perform the operation of obtaining the fitting height of the target object according to the point cloud of the target object, it is specifically configured to obtain the maximum fitting height of the target object. Height and fitting minimum height; obtaining the fitting height according to the fitting maximum height and the fitting minimum height.
- the point cloud sensor in this embodiment can be used to implement the technical solutions in the foregoing method embodiments, and its implementation principles and technical effects are similar, and will not be repeated here.
- FIG. 10 is a schematic structural diagram of a mobile platform provided by an embodiment of this application.
- the mobile platform of this embodiment includes: a point cloud sensor 101 and a processor 102.
- the point cloud sensor 101 is used to collect the point cloud of the target object; wherein, the point cloud sensor 101 is mounted on the mobile platform;
- the processor 102 is in communication connection with the point cloud sensor 101, and is configured to perform the following operations: generate a two-dimensional point cloud in the top view according to the point cloud of the target object collected by the point cloud sensor; and determine the two-dimensional point cloud In the visible area facing the mobile platform; according to the visible area, the fitting feature of the target object is determined, and the target object is fitted according to the fitting feature.
- the processor 102 when configured to perform the operation of determining the fitting feature of the target object according to the visible area, it is specifically configured to: determine according to the visible area and the two-dimensional point cloud The fitting feature.
- the fitting feature is a fitting angle of the target object
- the processor 102 when the processor 102 is configured to perform an operation of determining the fitting feature based on the visible area and the two-dimensional point cloud , Specifically used for: determining the fitting angle according to the visible area, the two-dimensional point cloud, and a preset angle range.
- the processor 102 when configured to perform the operation of determining the fitting angle according to the visible area, the two-dimensional point cloud, and a preset angle range, it is specifically configured to: according to the visible area , The two-dimensional point cloud and the preset angle range determine the corresponding fitting angle with the smallest visible edge cost.
- the processor 102 when the processor 102 is configured to perform the operation of determining the corresponding fitting angle with the least visible edge cost based on the visible area, the two-dimensional point cloud, and a preset angle range, specifically use In: selecting at least one reference angle from the preset angle range; and determining the corresponding fitting angle with the least visible edge cost according to the at least one reference angle, the visible area, and the two-dimensional point cloud.
- the processor 102 is configured to determine that the corresponding visible edge cost is the smallest according to the at least one reference angle, the visible area, and the two-dimensional point cloud.
- the fitting angle it is specifically used to: for any one of the multiple reference angles, according to the visible area and the two-dimensional point cloud, in the first reference angle Determine a first preselected angle with the smallest visible edge cost within a neighborhood range, where the first neighborhood range includes the first reference angle; determine the corresponding preselected angle with the least visible edge cost among the preselected angles as the fitting angle.
- the processor 102 is configured to perform determining a first preselected angle with the least visible edge cost within a first neighborhood range of the first reference angle based on the visible area and the two-dimensional point cloud During the operation, it is specifically used to obtain the first visible edge cost corresponding to the first reference angle according to the visible area and the two-dimensional point cloud, and update the first reference angle to the first neighbor According to the first angle within the domain, the second visible edge cost corresponding to the first angle is obtained according to the visible area and the two-dimensional point cloud; according to the second visible edge cost and the first visible edge The relationship between the costs, update the first angle to the second angle in the first neighborhood; repeat the execution of obtaining the visible edge costs corresponding to the angles in the first neighborhood, according to two adjacent The obtained relationship of the visible edge cost is an operation of updating the angle within the first neighborhood range until the first preselected angle with the smallest visible edge cost is obtained.
- the processor 102 is configured to determine that the corresponding visible edge cost is the smallest according to the at least one reference angle, the visible area, and the two-dimensional point cloud.
- the fitting angle it is specifically used to: determine the corresponding target reference angle with the least visible edge cost among multiple reference angles according to the visible area and the two-dimensional point cloud; according to the visible area And the two-dimensional point cloud, determining the corresponding fitting angle with the smallest visible edge cost within a second neighborhood range of the target reference angle, and the second neighborhood range includes the target reference angle.
- the processor 102 is configured to determine the fitting angle with the least visible edge cost within the second neighborhood range of the target reference angle according to the visible area and the two-dimensional point cloud During the operation, it is specifically used to: obtain multiple sub-reference angles from within the second neighborhood of the target reference angle; determine the visible edge cost corresponding to each sub-reference angle according to the visible area and the two-dimensional point cloud The smallest sub-reference angle is the fitting angle.
- the processor 102 when configured to perform the operation of obtaining the first visible edge cost corresponding to the first reference angle according to the visible area and the two-dimensional point cloud, it is specifically configured to: The first outer envelope rectangle of the two-dimensional point cloud in the first direction, the visible area and the first boundary and the second boundary corresponding to the visible area, to obtain the first visible edge cost; wherein, The first direction is a direction corresponding to the rotation of the first reference angle from the coordinate origin of the point cloud sensor coordinate system, and the coordinate origin of the point cloud sensor coordinate system is the position of the point cloud sensor.
- the processor 102 is configured to execute the first outer envelope rectangle in the first direction according to the two-dimensional point cloud, the visible area and the first boundary and the second boundary corresponding to the visible area.
- Boundary when the operation of obtaining the first visible edge cost is specifically used to: obtain the first area of the first area defined by the visible area, the first boundary, and the second boundary; The outer envelope rectangle faces the visible side of the mobile platform, and the second area of the second area defined by the first boundary and the second boundary; obtains the side of the first envelope rectangle outside the target area, the A third area of a third area defined by the first boundary and the second boundary, the target area is the area between the first boundary and the second boundary, and the target area includes the two-dimensional point cloud ; According to the first area, the second area and the third area, determine the first visible edge cost.
- the processor 102 is configured to perform an operation of determining the first visible edge cost corresponding to the first reference angle according to the first area, the second area, and the third area , It is specifically used to: obtain the first area, the second area, and the first preset weight, obtain the first preselected visible edge cost; according to the third area and the second preset weight, obtain the second preselect Visible edge cost; determining the first visible edge cost according to the first preselected visible edge cost and the second preselected visible edge cost.
- the angle included in the preset angle range is an angle between a first preset angle and a second preset angle, and the first preset angle minus the second preset angle is equal to 90°,
- the preset angle range includes the first preset angle and/or the second preset angle.
- the fitting feature includes at least one of the following: the fitting angle of the target object, the fitting direction of the target object, the fitting bounding box of the target object, and the fitting of the target object.
- the processor 102 when configured to perform the operation of fitting the target object according to the fitting feature, it is specifically configured to: obtain the information of the target object according to the point cloud of the target object Fitting height; fitting the target object according to the fitting feature and the fitting height.
- the processing when the fitting feature includes at least one of the following: the fitting angle of the target object, the fitting direction of the target object, and the fitting visible edge of the target object, the processing When the device 102 is used to perform the operation of fitting the target object according to the fitting feature and the fitting height, it is specifically used to: obtain the fitting boundary of the target object according to the fitting feature Box; according to the fitting bounding box and fitting height, fitting the target object.
- the processor 102 when configured to perform the operation of obtaining the fitting height of the target object according to the point cloud of the target object, it is specifically configured to obtain the maximum fitting height of the target object. Height and fitting minimum height; obtaining the fitting height according to the fitting maximum height and the fitting minimum height.
- the mobile platform of this embodiment can be used to execute the technical solutions in the foregoing method embodiments, and its implementation principles and technical effects are similar, and will not be repeated here.
- An embodiment of the present application also provides a computer-readable storage medium, including a program or instruction, and when the program or instruction runs on a computer, the method described in the foregoing method embodiment is executed.
- the aforementioned program can be stored in a computer readable storage medium.
- the program executes the steps including the above-mentioned method embodiments; and the aforementioned storage medium includes: ROM, RAM, magnetic disk, or optical disk and other media that can store program code.
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Multimedia (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Optical Radar Systems And Details Thereof (AREA)
- Length Measuring Devices By Optical Means (AREA)
Abstract
一种目标物体的拟合方法、点云传感器和移动平台,该方法包括:根据点云传感器采集的目标物体的点云,生成俯视图下的二维点云,其中点云传感器用于搭载于移动平台(S101);确定二维点云中朝向移动平台的可见区域(S102);根据目标物体的二维点云中朝向移动平台的可见区域,确定目标物体的拟合特征,并根据拟合特征对目标物体进行拟合(S103)。该方法可提高拟合目标物体的准确度。
Description
本申请涉及自动驾驶领域,尤其涉及一种目标物体的拟合方法、点云传感器和移动平台。
在自动驾驶任务中,基于车辆点云的目标车辆的识别、跟踪是一项重要的感知任务。车辆点云是一组离散点,实际处理中往往需要对判定为目标车辆的点云进行拟合,得到拟合后的车辆,以根据拟合后的车辆来判断其车辆的朝向和尺寸,进而完成跟踪任务。
通常用的拟合车辆的方法包括最小面积外包络矩形方法,该方法将车辆的三维点云投影至二维平面,得到二维点云,确定二维点云的面积最小的外包络矩形为拟合特征,根据该拟合特征得到拟合后的车辆。但是该方法对车辆的拟合均不够精确。
发明内容
本申请实施例提供一种目标物体的拟合方法、点云传感器和移动平台,提高了对目标物体拟合的精确度。
第一方面,本申请实施例提供一种目标物体的拟合方法,包括:根据点云传感器采集的目标物体的点云,生成俯视图下的二维点云,其中所述点云传感器用于搭载于移动平台;确定所述二维点云中朝向所述移动平台的可见区域;根据所述可见区域,确定所述目标物体的拟合特征,并根据所述拟合特征对所述目标物体进行拟合。
结合第一方面,在第一方面的一种可能的实现方式中,根据所述可见区域,确定所述目标物体的拟合特征,包括:根据所述可见区域和所述二维点云,确定所述拟合特征。
结合第一方面,在第一方面的一种可能的实现方式中,所述拟合特征为所述目标物体的拟合角度,所述根据所述可见区域和所述二维点云,确定 所述拟合特征,包括:根据所述可见区域、所述二维点云和预设角度范围,确定所述拟合角度。
结合第一方面,在第一方面的一种可能的实现方式中,所述根据所述可见区域、所述二维点云和预设角度范围,确定所述拟合角度,包括:根据所述可见区域、所述二维点云和预设角度范围,确定对应的可见边缘代价最小的所述拟合角度。
结合第一方面,在第一方面的一种可能的实现方式中,所述根据所述可见区域、所述二维点云和预设角度范围,确定对应的可见边缘代价最小的所述拟合角度,包括:从所述预设角度范围内选取至少一个参考角度;根据所述至少一个参考角度、所述可见区域和所述二维点云,确定对应的可见边缘代价最小的所述拟合角度。
结合第一方面,在第一方面的一种可能的实现方式中,在参考角度的数目为多个时,所述根据所述至少一个参考角度、所述可见区域和所述二维点云,确定对应的可见边缘代价最小的所述拟合角度,包括:对于多个参考角度中的任意一个第一参考角度,根据所述可见区域和所述二维点云,在所述第一参考角度的第一邻域范围内确定可见边缘代价最小的第一预选角度,所述第一邻域范围包括所述第一参考角度;确定各预选角度中对应的可见边缘代价最小的预选角度为所述拟合角度。
结合第一方面,在第一方面的一种可能的实现方式中,根据所述可见区域和所述二维点云,在所述第一参考角度的第一邻域范围内确定可见边缘代价最小的第一预选角度,包括:根据所述可见区域和所述二维点云,获取所述第一参考角度对应的第一可见边缘代价,并更新所述第一参考角度至所述第一邻域范围内的第一角度,根据所述可见区域和所述二维点云,获取所述第一角度对应的第二可见边缘代价;根据所述第二可见边缘代价和所述第一可见边缘代价之间的关系,更新所述第一角度至所述第一邻域范围内的第二角度;重复执行获取所述第一邻域范围内的角度对应的可见边缘代价、根据相邻两次获取到的可见边缘代价的关系在所述第一邻域范围内更新角度的操作,直至获取到可见边缘代价最小的所述第一预选角度。
结合第一方面,在第一方面的一种可能的实现方式中,在参考角度的数目为多个时,所述根据所述至少一个参考角度、所述可见区域和所述二维 点云,确定对应的可见边缘代价最小的所述拟合角度,包括:根据所述可见区域和所述二维点云,在多个参考角度中确定对应的可见边缘代价最小的目标参考角度;根据所述可见区域和所述二维点云,在所述目标参考角度的第二邻域范围内确定对应的可见边缘代价最小的所述拟合角度,所述第二邻域范围包括所述目标参考角度。
结合第一方面,在第一方面的一种可能的实现方式中,根据所述可见区域和所述二维点云,在所述目标参考角度的第二邻域范围内确定可见边缘代价最小的所述拟合角度,包括:从所述目标参考角度的第二邻域范围内获取多个子参考角度;根据所述可见区域和所述二维点云,确定各子参考角度对应的可见边缘代价中最小的子参考角度为所述拟合角度。
结合第一方面,在第一方面的一种可能的实现方式中,根据所述可见区域和所述二维点云,获取所述第一参考角度对应的第一可见边缘代价,包括:根据所述二维点云在第一方向上的第一外包络矩形、所述可见区域和所述可见区域对应的第一边界和第二边界,获取所述第一可见边缘代价;其中,所述第一方向为从点云传感器坐标系的坐标原点出发旋转所述第一参考角度后所对应的方向,所述点云传感器坐标系的坐标原点为所述点云传感器所在的位置。
结合第一方面,在第一方面的一种可能的实现方式中,所述根据所述二维点云在第一方向上的第一外包络矩形、所述可见区域和所述可见区域对应的第一边界和第二边界,获取所述第一可见边缘代价,包括:获取由所述可见区域、所述第一边界和第二边界限定的第一区域的第一面积;获取由所述第一外包络矩形朝向所述移动平台的可见边、所述第一边界和第二边界限定的第二区域的第二面积;获取所述第一外包络矩形位于目标区域外的边、所述第一边界和第二边界限定的第三区域的第三面积,所述目标区域为所述第一边界和第二边界之间的区域,所述目标区域内包括所述二维点云;根据所述第一面积、所述第二面积和所述第三面积,确定所述第一可见边缘代价。
结合第一方面,在第一方面的一种可能的实现方式中,所述根据所述第一面积、所述第二面积和所述第三面积,确定所述第一参考角度对应的所述第一可见边缘代价,包括:获取所述第一面积、所述第二面积和第一预设权重,获取第一预选可见边缘代价;根据所述第三面积和第二预设权重,获 取第二预选可见边缘代价;根据所述第一预选可见边缘代价和所述第二预选可见边缘代价,确定所述第一可见边缘代价。
结合第一方面,在第一方面的一种可能的实现方式中,所述预设角度范围包括的角度为第一预设角度与第二预设角度之间的角度,所述第一预设角度减去所述第二预设角度等于90°,所述预设角度范围包括所述第一预设角度和/或所述第二预设角度。
结合第一方面,在第一方面的一种可能的实现方式中,所述拟合特征包括如下中的至少一项:所述目标物体的拟合角度、所述目标物体的拟合方向、所述目标物体拟合边界框、所述目标物体的拟合可见边;
其中,所述拟合方向为从点云传感器坐标系的坐标原点出发旋转所述拟合角度后所对应的方向,所述点云传感器坐标系的坐标原点为所述点云传感器所在的位置。
结合第一方面,在第一方面的一种可能的实现方式中,所述根据所述拟合特征对所述目标物体进行拟合,包括:
根据所述目标物体的点云,获取所述目标物体的拟合高度;
根据所述拟合特征和所述拟合高度,对所述目标物体进行拟合。
结合第一方面,在第一方面的一种可能的实现方式中,在所述拟合特征包括如下中的至少一项时:所述目标物体的拟合角度、所述目标物体的拟合方向、所述目标物体的拟合可见边,根据所述拟合特征和所述拟合高度,对所述目标物体进行拟合,包括:
根据所述拟合特征,获取所述目标物体拟合边界框;
根据所述拟合边界框和拟合高度,对所述目标物体进行拟合。
结合第一方面,在第一方面的一种可能的实现方式中,根据所述目标物体的点云,获取所述目标物体的拟合高度,包括:
获取所述目标物体的点云拟合最大高度和拟合最小高度;
根据所述拟合最大高度和拟合最小高度,获取所述拟合高度。
第二方面,本申请实施例提供一种点云传感器,该点云传感器用于搭载在移动平台上,该点云传感器包括:
采集器,用于采集所述目标物体的点云;
处理器,与所述采集器通信连接,用于执行第一方面以及第一方面任 一可能的实现方式所述的方法。可以理解的是,在第二方面中,“根据点云传感器采集的目标物体的点云,生成俯视图下的二维点云”为“根据点云传感器的采集器采集的目标物体的点云,生成俯视图下的二维点云”。
第三方面,本申请实施例提供一种移动平台,包括:点云传感器,用于采集所述目标物体的点云;其中,所述点云传感器搭载于所述移动平台;处理器,与所述点云传感器通信连接,用于执行第一方面以及第一方面任一可能的实现方式所述的方法。
第四方面,本申请实施例提供一种移动平台,其特征在于,所述移动平台上搭载有第二方面所述的点云传感器。
第五方面,本申请实施例提供一种计算机可读存储介质,包括程序或指令,当所述程序或指令在计算机上运行时,第一方面以及第一方面任一可能的实现方式所述的方法被执行。
第六方面,本发明实施例提供一种计算机可读存储介质,该计算机可读存储介质存储有计算机程序,该计算机程序包含至少一段代码,该至少一段代码可由计算机执行,以控制所述计算机执行上述第一方面以及第一方面任一可能的实现方式所述的方法。
第七方面,本发明实施例提供一种计算机程序,当所述计算机程序被计算机执行时,用于执行上述第一方面以及第一方面任一可能的实现方式所述的方法。
所述程序可以全部或者部分存储在与处理器封装在一起的存储介质上,也可以部分或者全部存储在不与处理器封装在一起的存储介质上。存储介质例如为存储器。
由于目标物体靠近点云传感器或者靠近搭载点云传感器的移动平台的一个或两个面能够被点云传感器观测到,背离点云传感器或者背离搭载点云传感器的移动平台的两个或三个面无法被点云传感器观测到,因此,本申请中基于目标物体的点云传感器能够观测的面对目标物体进行拟合,即根据目标物体的二维点云中朝向移动平台的可见区域,对目标物体进行拟合,可以提高对目标物体拟合的准确性。
为了更清楚地说明本申请实施例或现有技术中的技术方案,下面将对实施例或现有技术描述中所需要使用的附图作一简单地介绍,显而易见地,下面描述中的附图是本申请的一些实施例,对于本领域普通技术人员来讲,在不付出创造性劳动性的前提下,还可以根据这些附图获得其他的附图。
图1为本申请实施例提供的应用场景示意图;
图2为本申请实施例提供的目标物体的拟合方法的流程图一;
图3为本申请实施例提供给的二维点云中朝向移动平台的可见区域的示意图;
图4为本申请实施提供的目标物体的拟合可见边的示意图一;
图5为本申请实施提供的目标物体的拟合可见边的示意图二;
图6为本申请实施提供的目标物体的拟合可见边的示意图三;
图7为本申请实施例提供的目标物体的拟合方法的流程图二;
图8为本申请实施例提供的获取可见边缘代价的示意图;
图9为本申请实施例提供的点云传感器的结构示意图;
图10为本申请实施例提供的移动平台的结构示意图。
为使本申请实施例的目的、技术方案和优点更加清楚,下面将结合本申请实施例中的附图,对本申请实施例中的技术方案进行清楚、完整地描述,显然,所描述的实施例是本申请一部分实施例,而不是全部的实施例。基于本申请中的实施例,本领域普通技术人员在没有作出创造性劳动前提下所获得的所有其他实施例,都属于本申请保护的范围。
下面结合附图,对本申请实施例进行说明。
图1为本申请实施例提供的应用场景示意图,参见图1,移动平台11上搭载有点云传感器12。
其中,点云传感器12用于采集目标物体的点云数据,目标物体的点云数据用于对目标物体进行拟合。
首先采用具体的实施例对本申请的目标物体的拟合方法进行说明。图2为本申请实施例提供的目标物体的拟合方法的流程图一。参见图2,本实施例的方法包括:
步骤S101、根据点云传感器采集的目标物体的点云,生成俯视图下的二维点云,其中,点云传感器用于搭载于移动平台。
具体地,点云传感器采集的目标物体的点云为三维点云,可将三维点云沿垂直方向投影到平面上得到二维点云,也就是根据目标物体的点云生成俯视图下的二维点云。例如,当目标物体为行驶在路面上的车辆时,可以将采集得到的车辆的三维点云沿垂直方向投影到路面的平面上,从而得到俯视图下的二维点云。
其中,本实施例中的点云传感器可以为TOF传感器,或者激光雷达。本实施例中的移动平台可为车辆,目标物体可为行驶在路上的其他车辆。在一个实施方式中,车辆上搭载有激光雷达,可以为一个或多个激光雷达,可以为旋转激光雷达或固态激光雷达,所述激光雷达用于获取所述车辆周围的三维点云信息。
步骤S102、确定二维点云中朝向移动平台的可见区域。
具体地,通常由于目标物体靠近点云传感器或者靠近搭载点云传感器的移动平台的一个或两个面能够被点云传感器观测到,背离点云传感器或者背离搭载点云传感器的移动平台的两个或三个面无法被点云传感器观测到,因此,为了拟合的准确性,本实施例基于目标物体的点云传感器能够观测到的面对目标物体进行拟合,即确定目标物体的二维点云中朝向移动平台的可见区域,以根据二维点云中朝向移动平台的可见区域,对目标物体进行拟合。其中,目标物体的二维点云中朝向移动平台的可见区域即为目标物体的二维点云在点云传感器的感知范围和视野范围内的区域。
图3为本申请实施例提供给的二维点云中朝向移动平台的可见区域的示意图。参见图3,图3中示意出了二维点云中的部分点。图3中所示的坐标原点O为点云传感器所在的位置,图3中的不规则图形21为目标物体的二维点云的轮廓或者凸壳,二维点云的轮廓或者凸壳中加粗部分211即为二维点云中朝向移动平台的可见区域,OA和OB为二维点云中朝向移动平台的可见区域对应的边界。
步骤S103、根据目标物体的二维点云中朝向移动平台的可见区域,确定目标物体的拟合特征,并根据拟合特征对目标物体进行拟合。
具体地,在确定了目标物体的二维点云中朝向移动平台的可见区域后, 即可根据该可见区域,确定目标物体的拟合特征,并根据拟合特征对目标物体进行拟合。
在一种方式中,根据拟合特征对目标物体进行拟合,包括:
a1、根据目标物体的点云,获取目标物体的拟合高度;
具体地,获取目标物体的点云的拟合最大高度和拟合最小高度,根据该拟合最大高度和该拟合最小高度,获取目标物体的拟合高度。其中,该拟合最大高度和拟合最小高度的差值即为目标物体的拟合高度。
其中,拟合最大高度可为目标物体的点云中高度最大的点的高度,也可为根据分层聚合得到的最大高度,比如根据具有一定厚度的点云中所有的点的高度拟合出来的,可以理解的是该一定厚度的点云包括目标物体的点云中高度最大的点。拟合最小高度可为目标物体的点云中高度最小的点的高度,也可为根据分层聚合得到的最小高度,比如根据具有一定厚度的点云中所有的点的高度拟合出来的,可以理解的是该一定厚度的点云包括目标物体的点云中高度最小的点。
a2、根据目标物体的拟合特征和目标物体的拟合高度,对目标物体进行拟合。
具体地,目标物体的拟合特征可包括如下中的至少一项:目标物体的拟合角度、目标物体的拟合方向、目标物体拟合边界框、目标物体的拟合可见边。目标物体的拟合特征还可为其它的特征,本实施例中并不限定。
在目标物体的拟合特征为目标物体的拟合角度的情况下,根据目标物体的拟合特征和目标物体的拟合高度,对目标物体进行拟合,包括:
b1、根据目标物体的拟合角度,获取目标物体的拟合方向;
具体地,确定从点云传感器坐标系的坐标原点出发旋转拟合角度后所对应的方向为目标物体的拟合方向,点云传感器坐标系的坐标原点为点云传感器所在的位置。
对于拟合角度的获取:根据目标物体的二维点云中朝向移动平台的可见区域、目标物体的二维点云和预设角度范围,确定目标物体的拟合角度,具体可为:根据目标物体的二维点云中朝向移动平台的可见区域、目标物体的二维点云和预设角度范围,在预选角度范围内确定目标物体的拟合角度。其中,预设角度范围和预选角度范围可相同也可不相同;预设角度范围包括的 角度为第一预设角度与第二预设角度之间的角度,第二预设角度减去第一预设角度等于90°,预设角度范围包括第一预设角度和/或第二预设角度;预选角度范围包括的角度为第三预设角度与第四预设角度之间的角度,第四预设角度减去第三预设角度等于90°,预选角度范围包括第三预设角度和/或第四预设角度。预选角度范围时根据预设角度范围得到的。
在一种方式中,目标物体的拟合角度为对应可见边缘代价最小的角度。具体地,每个角度均对应一个可见边缘代价。对于任意的一个角度:可见边缘代价为用于综合评价目标物体的二维点云在该角度对应的方向上的外包络矩形与该二维点云的贴合度以及该外包络矩形溢出目标区域的溢出度的参数;其中,该角度对应的方向为从点云传感器坐标系的坐标原点出发旋转该角度后所对应的方向。该目标区域为目标物体的二维点云中朝向移动平台的可见区域对应的第一边界和第二边界之间的区域,且该目标区域内包括目标物体的二维点云;进一步地,该第一边界和该第二边界能够包围该二维点云且该第一边界和该第二边界均与该二维点云的凸壳具有交点。继续参见图3,图3中的OA为目标物体的二维点云中朝向移动平台的可见区域对应的第一边界,OB为目标物体的二维点云中朝向移动平台的可见区域对应的第二边界。目标物体的二维点云在该角度对应的方向上的外包络矩形满足如下条件:该外包络矩形的两条边所在的方向与该角度对应的方向相同,该外包络矩形的另外两条边所在的方向与该角度对应的方向垂直。
其中,角度对应的可见边缘代价越小,在从点云传感器坐标系的坐标原点出发旋转该角度后所对应的方向上得到的目标物体的二维点云的外包络矩形与目标物体的横截面越接近,因此,目标物体的拟合角度为对应可见边缘代价最小的角度。
对于“根据目标物体的二维点云中朝向移动平台的可见区域、目标物体的二维点云和预设角度范围,确定目标物体的拟合角度”的具体实现,将在后续的实施例中进行详细阐述。
b2、在目标物体的拟合方向上,获取目标物体的拟合边界框。
具体地,获取目标物体的二维点云在拟合方向上的外包络矩形,该外包络矩形即为目标物体的拟合边界框。同样地,目标物体的二维点云在拟合方向上的外包络矩形满足如下条件:该外包络矩形的两条边所在的方向与该目 标物体的拟合方向相同,该外包络矩形的另外两条边所在的方向与该目标物体的拟合方向垂直。
在目标物体的拟合方向上,获取目标物体的拟合边界框的过程中,可先根据拟合方向,获取目标物体的拟合可见边,再以目标物体的拟合可见边为边得到目标物体的二维点云在拟合方向上的外包络矩形。
具体地,目标物体的拟合可见边可称为目标物体的L型特征或I型特征。可以理解的是,目标物体的拟合可见边满足以下条件:(1)目标物体的二维点云位于目标物体的拟合可见边的同一侧,且目标物体的二维点云的凸壳与目标物体的拟合可见边具有交点;(2)在目标物体的拟合可见边具有两条的情况下,两条拟合可见边中的一条边所在的方向与该目标物体的拟合方向相同,另一条边所在的方向与该目标物体的拟合方向垂直;在目标物体的拟合可见边具有一条的情况下,该拟合可见边所在的方向与该目标物体的拟合方向相同。
图4为本申请实施提供的目标物体的拟合可见边的示意图一,图5为本申请实施提供的目标物体的拟合可见边的示意图二,图6为本申请实施提供的目标物体的拟合可见边的示意图三。
参见图4,41和42为目标物体的拟合可见边,此时的拟合可见边为目标物体的L型特征。参见图5,51为目标物体的拟合可见边,此时的拟合可见边为目标物体的I型特征。参见图6,61为目标物体的拟合可见边,此时的拟合可见边为目标物体的I型特征。
b3、根据目标物体的拟合边界框和目标物体拟合高度,对目标物体进行拟合。
具体地,获取以横截面为目标物体的拟合边界框,高度为目标物体的拟合高度的立方体,该立方体即为拟合后的目标物体。
在目标物体的拟合特征为目标物体的拟合方向的情况下,根据目标物体的二维点云中朝向移动平台的可见区域,获取目标物体的拟合特征,包括:
c1、根据目标物体的二维点云中朝向移动平台的可见区域,确定根据目标物体的拟合角度。
c2、根据目标物体的拟合角度,确定目标物体的拟合方向。
具体地,c1~c2的具体实现参照b1中的阐述,本实施例中不在赘述
相应地,在该种情况下,根据目标物体的拟合特征和目标物体的拟合高度,对目标物体进行拟合,包括:
d1、根据目标物体的拟合方向,获取目标物体的拟合边界框;
具体地,该步骤的具体实现参照b2中的阐述,本实施例中不在赘述。
d2、根据目标物体的拟合边界框和目标物体拟合高度,对目标物体进行拟合。
具体地,该步骤的具体实现参照b3中的阐述,本实施例中不在赘述。
在目标物体的拟合特征为目标物体的拟合边界框的情况下,根据目标物体的二维点云中朝向移动平台的可见区域,确定目标物体的拟合特征,包括:
e1、根据目标物体的二维点云中朝向移动平台的可见区域,确定根据目标物体的拟合角度。
e2、根据目标物体的拟合角度,确定根据目标物体的拟合方向;
e3、在目标物体的拟合方向上,获取目标物体的拟合边界框。
相应地,在该情况下,根据目标物体的拟合特征和目标物体的拟合高度,对目标物体进行拟合,包括:根据目标物体的拟合边界框和目标物体的拟合高度,对目标物体进行拟合。
在目标物体的拟合特征为目标物体的拟合可见边的情况下,根据目标物体的二维点云中朝向移动平台的可见区域,获取目标物体的拟合特征,包括:
f1、根据目标物体的二维点云中朝向移动平台的可见区域,确定根据目标物体的拟合角度。
f2、根据目标物体的拟合角度,确定根据目标物体的拟合方向;
f3、根据目标物体的拟合方向,获取目标物体的拟合可见边。
相应地,在该种情况下,根据目标物体的拟合特征和目标物体的拟合高度,对目标物体进行拟合,包括:
g1、根据目标物体的拟合可见边,获取目标物体的拟合边界框;
g2、根据目标物体的拟合边界框和目标物体的拟合高度,对目标物体进行拟合。
由于目标物体靠近点云传感器或者靠近搭载点云传感器的移动平台的一个或两个面能够被点云传感器观测到,背离点云传感器或者背离搭载点云传感器的移动平台的两个或三个面无法被点云传感器观测到,因此,本实施例 中基于目标物体的点云传感器能够观测到的面对目标物体进行拟合,即根据目标物体的二维点云中朝向移动平台的可见区域,对目标物体进行拟合,可以提高对目标物体拟合的准确性。
下面采用具体的实施例对上一实施例中的“根据目标物体的二维点云中朝向移动平台的可见区域、该二维点云和预设角度范围,确定目标物体的拟合角度”的一种可实现的方式进行详细说明。
图7为本申请实施例提供的目标物体的拟合方法的流程图二,参见图7,本实施例的方法包括:
步骤S201、从预设角度范围内选取至少一个参考角度。
具体地,本实施例中的预设角度范围的含义与上一实施例中的预设角度范围的含义相同。
其中,至少一个参考角度是指一个或多个参考角度。在参考角度为一个时,参考角度可为第一预设参考范围内的任一角度。示例性地,在预设角度范围为[0°,90°]时,一个参考角度可为45°。
在参考角度为多个时,可在预设角度范围内每间隔相同的角度选取一个参考角度,得到多个参考角度。示例性地,在预设角度范围为[0°,90°]时,多个参考角度可包括:0°、10°、20°、30°、40°、50°、60°、70°、80°、90°,在预设角度范围为[0°,90°)时,多个参考角度可包括:0°、10°、20°、30°、40°、50°、60°、70°、80°。
步骤S202、根据至少一个参考角度、目标物体的二维点云中朝向移动平台的可见区域和该二维点云,确定对应的可见边缘代价最小的拟合角度。
具体地,下面以多个参考角度为例来说明“根据多个参考角度和该可见区域,确定对应的可见边缘代价最小的拟合角度”的方法。
在第一种实现方式中:“根据多个参考角度和该可见区域,确定对应的可见边缘代价最小的拟合角度”可通过如下的步骤h1~h2实现。
h1、对于多个参考角度中的任意一个第一参考角度,根据目标物体的二维点云中朝向移动平台的可见区域,在第一参考角度的第一邻域范围内确定可见边缘代价最小的第一预选角度,第一邻域范围内包括第一参考角度。
具体地,每个参考角度具有一个邻域范围,因此具有多个邻域范围,多个邻域范围的并集为上一实施例中的预选角度范围。
在预设角度范围为[P
1°,P
2°)或[P
1°,P
2°]或(P
1°,P
2°],且在预设角度范围内每间隔α°选取一个参考角度的情况下,在一种方式中,若第一参考角度为N°,第一邻域范围为(N°-α°/2,N°+α°/2]或[N°-α°/2,N°+α°/2),此时,预设角度范围和预选角度范围不相同;在另一种方式中,若第一参考角度为N°,在N不等于P
1也不等于P
2时,第一邻域范围为(N°-α°/2,N°+α°/2]或[N°-α°/2,N°+α°/2),在N等于P
1时,第一邻域范围为[N°,N°+α°/2]或(N°,N°+α°/2]或[N°,N°+α°/2)或(N°,N°+α°/2),在N等于P
2时,第一邻域范围为(N°-α°/2,N°)或(N°-α°/2,N°]或[N°-α°/2,N°)或[N°-α°/2,N°],此时,预设角度范围和预选角度范围可能相同
可采用梯度下降法,在第一参考角度的第一邻域范围内确定可见边缘代价最小的第一预选角度,具体为:
j1、根据该可见区域和目标物体的二维点云,获取第一参考角度对应的第一可见边缘代价。
具体地,获取第一可见边缘代价的方法具体如下:根据二维点云在第一方向上的第一外包络矩形、该可见区域和该可见区域对应的第一边界和第二边界,获取第一参考角度对应的第一可见边缘代价;其中,第一方向为从点云传感器坐标系的坐标原点出发旋转第一参考角度后所对应的方向。该可见区域对应的第一边界和第二边界同上一实施例中的阐述。
图8为本申请实施例提供的获取可见边缘代价的示意图,下面结合图8对“根据二维点云在第一方向上的第一外包络矩形、该可见区域和该可见区域对应的第一边界和第二边界,获取第一参考角度对应的第一可见边缘代价”的过程进行说明。具体地,“根据二维点云在第一方向上的第一外包络矩形、该可见区域和该可见区域对应的第一边界和第二边界,获取第一参考角度对应的第一可见边缘代价”,包括:
j11、获取由该可见区域、第一边界和第二边界限定的第一区域的第一面积。
具体地,图8中的OA为第一边界,OB为第二边界,第一区域即为图8中的区域81。
j12、获取由该第一外包络矩形朝向移动平台的可见边、第一边界和第二边界限定的第二区域的第二面积。
具体地,第一方向为图8中箭头所指示的方向,第二区域为图8中的区域82。
j13、获取第一外包络矩形位于目标区域外的边、该第一边界和该第二边界限定的第三区域的第三面积,该目标区域为该第一边界和该第二边界之间的区域且该目标区域内包括目标物体的二维点云。
具体地,第三区域为图8中的区域83,也就是实线填充的区域。
j14、根据第一面积、第二面积和第三面积,确定第一参考角度对应的可见边缘代价。
具体地,根据第一面积、第二面积和第三面积,确定第一参考角度对应的可见边缘代价,包括:
j141、根据第一面积、第二面积和第一预设权重,获取第一预选可见边缘代价。
具体地,第一预选可见边缘代价用于表征第一外包络矩形与二维点云的凸壳的贴合度。其中,第一预选可见边缘代价等于第一面积与第二面积的差值乘以第一预设权重。第一面积与第二面积的差值为图中虚线填充的区域的面积。
第一预选可见边缘代价的计算过程可表示为:cost1=α×(S
1-S
2),其中,cost1为第一预选可见边缘代价,α为第一预设权重,S
1为第一面积,S
2为第二面积。α可为0.5。
j142、根据第三面积和第二预设权重,获取第二预选可见边缘代价。
具体地,第二预选可见边缘代价用于表征第一外包络矩形溢出目标区域的溢出度。其中,第二预选可见边缘代价等于第三面积乘以第二预设权重。
第二预选可见边缘代价的计算过程可表示为:cost2=β×S
3,其中,cost2为第二预选可见边缘代价,β为第二预设权重,S
3为第三面积,β可为0.5,α+β=1。
j143、根据第一预选可见边缘代价和第二预选可见边缘代价,确定第一参考角度对应的第一可见边缘代价。
具体地,第一预选可见边缘代价和第二预选可见边缘代价之和为第一参考角度对应的第一可见边缘代价。可表示为:cost=cost1+cost2,其中,cost为第一可见边缘代价。
因此,可见边缘代价是综合目标物体的二维点云在某一角度对应的方向上的外包络矩形与该二维点云的贴合度以及该外包络矩形溢出目标区域的溢出度的参数。
j2、更新第一参考角度至第一邻域范围内的第一角度,并根据该可见区域和目标物体的二维点云,获取该第一角度对应的第二可见边缘代价。
具体地,第一角度可为大于第一参考角度的角度,也可为小于第一参考角度的角度。
此处,获取该第一角度对应的第二可见边缘代价的方法与获取第一参考角度对应的第一可见边缘代价的方法相同,此处不再赘述。
j3、根据该第二可见边缘代价和该第一可见边缘代价之间的关系,更新该第一角度至第一邻域范围内的第二角度。
具体地,角度更新规则如下:
若第一角度大于第一参考角度,且第一可见边缘代价大于第二可见边缘代价,则更新该第一角度至第一邻域范围内比第一角度小的第二角度。
若第一角度大于第一参考角度,且第一可见边缘代价小于第二可见边缘代价,则更新该第一角度至第一邻域范围内比第一角度大的第二角度。
若第一角度小于第一参考角度,且第一可见边缘代价大于第二可见边缘代价,则更新该第一角度至第一邻域范围内比第一角度大的第二角度。
若第一角度小于第一参考角度,且第一可见边缘代价小于第二可见边缘代价,则更新该第一角度至第一邻域范围内比第一角度小的第二角度。
更新该第一角度至第一邻域范围内的第二角度后,重复执行获取第一邻域范围内的角度对应的可见边缘代价、根据相邻两次获取到的可见边缘代价的关系在第一邻域范围内更新角度的操作,直至获取到可见边缘代价最小的第一预选角度。
如上的j1~j3为采用梯度下降法在第一参考角度的第一邻域范围内确定可见边缘代价最小的第一预选角度的具体实现。
对于多个参考角度中的其它参考角度,均按照获取第一参考角度对应的第一预选角度的方法,获取其它参考角度对应的预选角度,最终得到多个预选角度。
其中,在采用梯度下降法获取目标物体的拟合角度时,选取多个参考角 度,可以防止限于局部最小值,提高了目标物体拟合的准确度。
h2、确定各预选角度中对应可见边缘代价最小的预选角度为目标物体的拟合角度。
根据至少一个参考角度、该可见区域和目标物体的二维点云,确定对应的可见边缘代价最小的目标物体的拟合角度的第一种实现方式,使得获取到的目标物体的拟合角度具有较高的准确度。
在第二种实现方式中:在参考角度的数目为多个时,根据至少一个参考角度、该可见区域和目标物体的二维点云,确定对应的可见边缘代价最小的目标物体的拟合角度,包括:
m1、根据可见区域和目标物体的二维点云,在多个参考角度中确定对应的可见边缘代价最小的目标参考角度。
具体地,获取每个参考角度对应的可见边缘代价,将对应的可见边缘代价最小的参考角度确定为目标参考角度。其中,获取每个参考角度对应的可见边缘代价参照获取第一参考角度对应的可见边缘代价的获取方法。
m2、根据该可见区域和目标物体的二维点云,在目标参考角度的第二邻域范围内确定对应的可见边缘代价最小的目标物体的拟合角度,第二邻域范围包括目标参考角度。
具体地,在目标参考角度为N°时,第二邻域范围为(S
1°,S
2°),其中,S
1°为与目标参考角度的差值的绝对值最小且小于目标参考角度的参考角度,S
2°为与目标参考角度的差值的绝对值最小且大于目标参考角度的参考角度。
在一种方式中,可采用梯度下降法,根据该可见区域和目标物体的二维点云,在目标参考角度的第二邻域范围内确定对应的可见边缘代价最小的目标物体的拟合角度。该方式的具体实现参照在第一参考角度的第一邻域范围内采用梯度下降法中确定第一预选角度的方法,此处不再赘述。
在另一种方式中,从目标参考角度的第二邻域范围内获取多个子参考角度,根据该可见区域和目标物体的二维点云,确定各子参考角度对应的可见边缘代价中最小的子参考角度为目标物体的拟合角度。
具体地,在该种方式中,从目标参考角度的第二邻域范围内获取多个子参考角度,获取每个子参考角度对应的可见边缘代价,将对应的可见边缘代价最小的子参考角度确定为目标物体的拟合角度。其中,获取每个子参考角 度对应的可见边缘代价参照获取第一参考角度对应的可见边缘代价的获取方法,此处不再赘述。
具体地,可在目标参考角度的第二邻域范围内每间隔相同的角度获取一个子参考角度。
根据至少一个参考角度、该可见区域和目标物体的二维点云,确定对应的可见边缘代价最小的目标物体的拟合角度的第二种实现方式,使得获取目标物体的拟合角度的效率比较高。
本实施例提供了获取目标物体的拟合角度的具体方法,本实施例的方法可进一步提高获取目标物体的拟合角度的准确度。
下面采用具体的实施例对实现上述目标物体的拟合方法的装置进行说明。
图9为本申请实施例提供的点云传感器的结构示意图,本实施例提供的点云传感器用于搭载在移动平台上,该点云传感器包括:采集器91和处理器92。
采集器91,用于采集所述目标物体的点云;
处理器92,与所述采集器通信连接,用于执行如下操作:根据所述采集器采集的目标物体的点云,生成俯视图下的二维点云;确定所述二维点云中朝向所述移动平台的可见区域;根据所述可见区域,确定所述目标物体的拟合特征,并根据所述拟合特征对所述目标物体进行拟合。
可选地,所述处理器92在用于执行根据所述可见区域,确定所述目标物体的拟合特征的操作时,具体用于:根据所述可见区域和所述二维点云,确定所述拟合特征。
可选地,所述拟合特征为所述目标物体的拟合角度,所述处理器92在用于执行根据所述可见区域和所述二维点云,确定所述拟合特征的操作时,具体用于:根据所述可见区域、所述二维点云和预设角度范围,确定所述拟合角度。
可选地,所述处理器92在用于执行根据所述可见区域、所述二维点云和预设角度范围,确定所述拟合角度的操作时,具体用于:根据所述可见区域、所述二维点云和预设角度范围,确定对应的可见边缘代价最小的所述拟合角度。
可选地,所述处理器92在用于执行根据所述可见区域、所述二维点云和 预设角度范围,确定对应的可见边缘代价最小的所述拟合角度的操作时,具体用于:从所述预设角度范围内选取至少一个参考角度;根据所述至少一个参考角度、所述可见区域和所述二维点云,确定对应的可见边缘代价最小的所述拟合角度。
可选地,在参考角度的数目为多个时,所述处理器92在用于执行根据所述至少一个参考角度、所述可见区域和所述二维点云,确定对应的可见边缘代价最小的所述拟合角度的操作时,具体用于:对于多个参考角度中的任意一个第一参考角度,根据所述可见区域和所述二维点云,在所述第一参考角度的第一邻域范围内确定可见边缘代价最小的第一预选角度,所述第一邻域范围包括所述第一参考角度;确定各预选角度中对应的可见边缘代价最小的预选角度为所述拟合角度。
可选地,所述处理器92在用于执行根据所述可见区域和所述二维点云,在所述第一参考角度的第一邻域范围内确定可见边缘代价最小的第一预选角度的操作时,具体用于:根据所述可见区域和所述二维点云,获取所述第一参考角度对应的第一可见边缘代价,并更新所述第一参考角度至所述第一邻域范围内的第一角度,根据所述可见区域和所述二维点云,获取所述第一角度对应的第二可见边缘代价;根据所述第二可见边缘代价和所述第一可见边缘代价之间的关系,更新所述第一角度至所述第一邻域范围内的第二角度;重复执行获取所述第一邻域范围内的角度对应的可见边缘代价、根据相邻两次获取到的可见边缘代价的关系在所述第一邻域范围内更新角度的操作,直至获取到可见边缘代价最小的所述第一预选角度。
可选地,在参考角度的数目为多个时,所述处理器92在用于执行根据所述至少一个参考角度、所述可见区域和所述二维点云,确定对应的可见边缘代价最小的所述拟合角度的操作时,具体用于:根据所述可见区域和所述二维点云,在多个参考角度中确定对应的可见边缘代价最小的目标参考角度;根据所述可见区域和所述二维点云,在所述目标参考角度的第二邻域范围内确定对应的可见边缘代价最小的所述拟合角度,所述第二邻域范围包括所述目标参考角度。
可选地,所述处理器92在用于执行根据所述可见区域和所述二维点云,在所述目标参考角度的第二邻域范围内确定可见边缘代价最小的所述拟合角 度的操作时,具体用于:从所述目标参考角度的第二邻域范围内获取多个子参考角度;根据所述可见区域和所述二维点云,确定各子参考角度对应的可见边缘代价中最小的子参考角度为所述拟合角度。
可选地,所述处理器92在用于执行根据所述可见区域和所述二维点云,获取所述第一参考角度对应的第一可见边缘代价的操作时,具体用于:根据所述二维点云在第一方向上的第一外包络矩形、所述可见区域和所述可见区域对应的第一边界和第二边界,获取所述第一可见边缘代价;其中,所述第一方向为从点云传感器坐标系的坐标原点出发旋转所述第一参考角度后所对应的方向,所述点云传感器坐标系的坐标原点为所述点云传感器所在的位置。
可选地,所述处理器92在用于执行根据所述二维点云在第一方向上的第一外包络矩形、所述可见区域和所述可见区域对应的第一边界和第二边界,获取所述第一可见边缘代价的操作时,具体用于:获取由所述可见区域、所述第一边界和第二边界限定的第一区域的第一面积;获取由所述第一外包络矩形朝向所述移动平台的可见边、所述第一边界和第二边界限定的第二区域的第二面积;获取所述第一外包络矩形位于目标区域外的边、所述第一边界和所述第二边界限定的第三区域的第三面积,所述目标区域为所述第一边界和所述第二边界之间的区域,所述目标区域内包括目标物体的二维点云;根据所述第一面积、所述第二面积和所述第三面积,确定所述第一可见边缘代价。
可选地,所述处理器92在用于执行根据所述第一面积、所述第二面积和所述第三面积,确定所述第一参考角度对应的所述第一可见边缘代价的操作时,具体用于:获取所述第一面积、所述第二面积和第一预设权重,获取第一预选可见边缘代价;根据所述第三面积和第二预设权重,获取第二预选可见边缘代价;根据所述第一预选可见边缘代价和所述第二预选可见边缘代价,确定所述第一可见边缘代价。
可选地,所述预设角度范围包括的角度为第一预设角度与第二预设角度之间的角度,所述第一预设角度减去所述第二预设角度等于90°,所述预设角度范围包括所述第一预设角度和/或所述第二预设角度。
可选地,所述拟合特征包括如下中的至少一项:所述目标物体的拟合角度、所述目标物体的拟合方向、所述目标物体拟合边界框、所述目标物体的 拟合可见边;其中,所述拟合方向为从点云传感器坐标系的坐标原点出发旋转所述拟合角度后所对应的方向,所述点云传感器坐标系的坐标原点为所述点云传感器所在的位置。
可选地,所述处理器92在用于执行根据所述拟合特征对所述目标物体进行拟合的操作时,具体用于:根据所述目标物体的点云,获取所述目标物体的拟合高度;根据所述拟合特征和所述拟合高度,对所述目标物体进行拟合。
可选地,在所述拟合特征包括如下中的至少一项时:所述目标物体的拟合角度、所述目标物体的拟合方向、所述目标物体的拟合可见边,所述处理器92在用于执行根据所述拟合特征和所述拟合高度,对所述目标物体进行拟合的操作时,具体用于:根据所述拟合特征,获取所述目标物体拟合边界框;根据所述拟合边界框和拟合高度,对所述目标物体进行拟合。
可选地,所述处理器92在用于执行根据所述目标物体的点云,获取所述目标物体的拟合高度的操作时,具体用于:获取所述目标物体的点云拟合最大高度和拟合最小高度;根据所述拟合最大高度和拟合最小高度,获取所述拟合高度。
本实施例的点云传感器,可以用于执行上述各方法实施例中的技术方案,其实现原理和技术效果类似,此处不再赘述。
图10为本申请实施例提供的移动平台的结构示意图,参见图10,本实施例的移动平台,包括:点云传感器101和处理器102。
点云传感器101,用于采集所述目标物体的点云;其中,所述点云传感器101搭载于所述移动平台;
处理器102,与所述点云传感器101通信连接,用于执行如下操作:根据所述点云传感器采集的目标物体的点云,生成俯视图下的二维点云;确定所述二维点云中朝向所述移动平台的可见区域;根据所述可见区域,确定所述目标物体的拟合特征,并根据所述拟合特征对所述目标物体进行拟合。
可选地,所述处理器102在用于执行根据所述可见区域,确定所述目标物体的拟合特征的操作时,具体用于:根据所述可见区域和所述二维点云,确定所述拟合特征。
可选地,所述拟合特征为所述目标物体的拟合角度,所述处理器102在用于执行根据所述可见区域和所述二维点云,确定所述拟合特征的操作时, 具体用于:根据所述可见区域、所述二维点云和预设角度范围,确定所述拟合角度。
可选地,所述处理器102在用于执行根据所述可见区域、所述二维点云和预设角度范围,确定所述拟合角度的操作时,具体用于:根据所述可见区域、所述二维点云和预设角度范围,确定对应的可见边缘代价最小的所述拟合角度。
可选地,所述处理器102在用于执行根据所述可见区域、所述二维点云和预设角度范围,确定对应的可见边缘代价最小的所述拟合角度的操作时,具体用于:从所述预设角度范围内选取至少一个参考角度;根据所述至少一个参考角度、所述可见区域和所述二维点云,确定对应的可见边缘代价最小的所述拟合角度。
可选地,在参考角度的数目为多个时,所述处理器102在用于执行根据所述至少一个参考角度、所述可见区域和所述二维点云,确定对应的可见边缘代价最小的所述拟合角度的操作时,具体用于:对于多个参考角度中的任意一个第一参考角度,根据所述可见区域和所述二维点云,在所述第一参考角度的第一邻域范围内确定可见边缘代价最小的第一预选角度,所述第一邻域范围包括所述第一参考角度;确定各预选角度中对应的可见边缘代价最小的预选角度为所述拟合角度。
可选地,所述处理器102在用于执行根据所述可见区域和所述二维点云,在所述第一参考角度的第一邻域范围内确定可见边缘代价最小的第一预选角度的操作时,具体用于:根据所述可见区域和所述二维点云,获取所述第一参考角度对应的第一可见边缘代价,并更新所述第一参考角度至所述第一邻域范围内的第一角度,根据所述可见区域和所述二维点云,获取所述第一角度对应的第二可见边缘代价;根据所述第二可见边缘代价和所述第一可见边缘代价之间的关系,更新所述第一角度至所述第一邻域范围内的第二角度;重复执行获取所述第一邻域范围内的角度对应的可见边缘代价、根据相邻两次获取到的可见边缘代价的关系在所述第一邻域范围内更新角度的操作,直至获取到可见边缘代价最小的所述第一预选角度。
可选地,在参考角度的数目为多个时,所述处理器102在用于执行根据所述至少一个参考角度、所述可见区域和所述二维点云,确定对应的可见边 缘代价最小的所述拟合角度的操作时,具体用于:根据所述可见区域和所述二维点云,在多个参考角度中确定对应的可见边缘代价最小的目标参考角度;根据所述可见区域和所述二维点云,在所述目标参考角度的第二邻域范围内确定对应的可见边缘代价最小的所述拟合角度,所述第二邻域范围包括所述目标参考角度。
可选地,所述处理器102在用于执行根据所述可见区域和所述二维点云,在所述目标参考角度的第二邻域范围内确定可见边缘代价最小的所述拟合角度的操作时,具体用于:从所述目标参考角度的第二邻域范围内获取多个子参考角度;根据所述可见区域和所述二维点云,确定各子参考角度对应的可见边缘代价中最小的子参考角度为所述拟合角度。
可选地,所述处理器102在用于执行根据所述可见区域和所述二维点云,获取所述第一参考角度对应的第一可见边缘代价的操作时,具体用于:根据所述二维点云在第一方向上的第一外包络矩形、所述可见区域和所述可见区域对应的第一边界和第二边界,获取所述第一可见边缘代价;其中,所述第一方向为从点云传感器坐标系的坐标原点出发旋转所述第一参考角度后所对应的方向,所述点云传感器坐标系的坐标原点为所述点云传感器所在的位置。
可选地,所述处理器102在用于执行根据所述二维点云在第一方向上的第一外包络矩形、所述可见区域和所述可见区域对应的第一边界和第二边界,获取所述第一可见边缘代价的操作时,具体用于:获取由所述可见区域、所述第一边界和第二边界限定的第一区域的第一面积;获取由所述第一外包络矩形朝向所述移动平台的可见边、所述第一边界和第二边界限定的第二区域的第二面积;获取所述第一外包络矩形位于目标区域外的边、所述第一边界和所述第二边界限定的第三区域的第三面积,所述目标区域为所述第一边界和第二边界之间的区域,所述目标区域内包括所述二维点云;根据所述第一面积、所述第二面积和所述第三面积,确定所述第一可见边缘代价。
可选地,所述处理器102在用于执行根据所述第一面积、所述第二面积和所述第三面积,确定所述第一参考角度对应的所述第一可见边缘代价的操作时,具体用于:获取所述第一面积、所述第二面积和第一预设权重,获取第一预选可见边缘代价;根据所述第三面积和第二预设权重,获取第二预选可见边缘代价;根据所述第一预选可见边缘代价和所述第二预选可见边缘代 价,确定所述第一可见边缘代价。
可选地,所述预设角度范围包括的角度为第一预设角度与第二预设角度之间的角度,所述第一预设角度减去所述第二预设角度等于90°,所述预设角度范围包括所述第一预设角度和/或所述第二预设角度。
可选地,所述拟合特征包括如下中的至少一项:所述目标物体的拟合角度、所述目标物体的拟合方向、所述目标物体拟合边界框、所述目标物体的拟合可见边;其中,所述拟合方向为从点云传感器坐标系的坐标原点出发旋转所述拟合角度后所对应的方向,所述点云传感器坐标系的坐标原点为所述点云传感器所在的位置。
可选地,所述处理器102在用于执行根据所述拟合特征对所述目标物体进行拟合的操作时,具体用于:根据所述目标物体的点云,获取所述目标物体的拟合高度;根据所述拟合特征和所述拟合高度,对所述目标物体进行拟合。
可选地,在所述拟合特征包括如下中的至少一项时:所述目标物体的拟合角度、所述目标物体的拟合方向、所述目标物体的拟合可见边,所述处理器102在用于执行根据所述拟合特征和所述拟合高度,对所述目标物体进行拟合的操作时,具体用于:根据所述拟合特征,获取所述目标物体拟合边界框;根据所述拟合边界框和拟合高度,对所述目标物体进行拟合。
可选地,所述处理器102在用于执行根据所述目标物体的点云,获取所述目标物体的拟合高度的操作时,具体用于:获取所述目标物体的点云拟合最大高度和拟合最小高度;根据所述拟合最大高度和拟合最小高度,获取所述拟合高度。
本实施例移动平台,可以用于执行上述各方法实施例中的技术方案,其实现原理和技术效果类似,此处不再赘述。
本申请实施例还提供一种计算机可读存储介质,包括程序或指令,当所述程序或指令在计算机上运行时,上述方法实施例所述的方法被执行。
本领域普通技术人员可以理解:实现上述各方法实施例的全部或部分步骤可以通过程序指令相关的硬件来完成。前述的程序可以存储于一计算机可读取存储介质中。该程序在执行时,执行包括上述各方法实施例的步骤;而前述的存储介质包括:ROM、RAM、磁碟或者光盘等各种可以存 储程序代码的介质。
最后应说明的是:以上各实施例仅用以说明本申请的技术方案,而非对其限制;尽管参照前述各实施例对本申请进行了详细的说明,本领域的普通技术人员应当理解:其依然可以对前述各实施例所记载的技术方案进行修改,或者对其中部分或者全部技术特征进行等同替换;而这些修改或者替换,并不使相应技术方案的本质脱离本申请各实施例技术方案的范围。
Claims (53)
- 一种目标物体的拟合方法,其特征在于,包括:根据点云传感器采集的目标物体的点云,生成俯视图下的二维点云,其中所述点云传感器用于搭载于移动平台;确定所述二维点云中朝向所述移动平台的可见区域;根据所述可见区域,确定所述目标物体的拟合特征,并根据所述拟合特征对所述目标物体进行拟合。
- 根据权利要求1所述的方法,其特征在于,根据所述可见区域,确定所述目标物体的拟合特征,包括:根据所述可见区域和所述二维点云,确定所述拟合特征。
- 根据权利要求2所述的方法,其特征在于,所述拟合特征为所述目标物体的拟合角度,所述根据所述可见区域和所述二维点云,确定所述拟合特征,包括:根据所述可见区域、所述二维点云和预设角度范围,确定所述拟合角度。
- 根据权利要求3所述的方法,其特征在于,所述根据所述可见区域、所述二维点云和预设角度范围,确定所述拟合角度,包括:根据所述可见区域、所述二维点云和预设角度范围,确定对应的可见边缘代价最小的所述拟合角度。
- 根据权利要求4所述的方法,其特征在于,所述根据所述可见区域、所述二维点云和预设角度范围,确定对应的可见边缘代价最小的所述拟合角度,包括:从所述预设角度范围内选取至少一个参考角度;根据所述至少一个参考角度、所述可见区域和所述二维点云,确定对应的可见边缘代价最小的所述拟合角度。
- 根据权利要求5所述的方法,其特征在于,在参考角度的数目为多个时,所述根据所述至少一个参考角度、所述可见区域和所述二维点云,确定对应的可见边缘代价最小的所述拟合角度,包括:对于多个参考角度中的任意一个第一参考角度,根据所述可见区域和所述二维点云,在所述第一参考角度的第一邻域范围内确定可见边缘代价最小的第一预选角度,所述第一邻域范围包括所述第一参考角度;确定各预选角度中对应的可见边缘代价最小的预选角度为所述拟合角度。
- 根据权利要求6所述的方法,其特征在于,根据所述可见区域和所述二维点云,在所述第一参考角度的第一邻域范围内确定可见边缘代价最小的第一预选角度,包括:根据所述可见区域和所述二维点云,获取所述第一参考角度对应的第一可见边缘代价,并更新所述第一参考角度至所述第一邻域范围内的第一角度,根据所述可见区域和所述二维点云,获取所述第一角度对应的第二可见边缘代价;根据所述第二可见边缘代价和所述第一可见边缘代价之间的关系,更新所述第一角度至所述第一邻域范围内的第二角度;重复执行获取所述第一邻域范围内的角度对应的可见边缘代价、根据相邻两次获取到的可见边缘代价的关系在所述第一邻域范围内更新角度的操作,直至获取到可见边缘代价最小的所述第一预选角度。
- 根据权利要求5所述的方法,其特征在于,在参考角度的数目为多个时,所述根据所述至少一个参考角度、所述可见区域和所述二维点云,确定对应的可见边缘代价最小的所述拟合角度,包括:根据所述可见区域和所述二维点云,在多个参考角度中确定对应的可见边缘代价最小的目标参考角度;根据所述可见区域和所述二维点云,在所述目标参考角度的第二邻域范围内确定对应的可见边缘代价最小的所述拟合角度,所述第二邻域范围包括所述目标参考角度。
- 根据权利要求8所述的方法,其特征在于,根据所述可见区域和所述二维点云,在所述目标参考角度的第二邻域范围内确定可见边缘代价最小的所述拟合角度,包括:从所述目标参考角度的第二邻域范围内获取多个子参考角度;根据所述可见区域和所述二维点云,确定各子参考角度对应的可见边缘代价中最小的子参考角度为所述拟合角度。
- 根据权利要求7所述的方法,其特征在于,根据所述可见区域和所述二维点云,获取所述第一参考角度对应的第一可见边缘代价,包括:根据所述二维点云在第一方向上的第一外包络矩形、所述可见区域和所 述可见区域对应的第一边界和第二边界,获取所述第一可见边缘代价;其中,所述第一方向为从点云传感器坐标系的坐标原点出发旋转所述第一参考角度后所对应的方向,所述点云传感器坐标系的坐标原点为所述点云传感器所在的位置。
- 根据权利要求10所述的方法,其特征在于,所述根据所述二维点云在第一方向上的第一外包络矩形、所述可见区域和所述可见区域对应的第一边界和第二边界,获取所述第一可见边缘代价,包括:获取由所述可见区域、所述第一边界和所述第二边界限定的第一区域的第一面积;获取由所述第一外包络矩形朝向所述移动平台的可见边、所述第一边界和第二边界限定的第二区域的第二面积;获取所述第一外包络矩形位于目标区域外的边、所述第一边界和第二边界限定的第三区域的第三面积,所述目标区域为所述第一边界和第二边界之间的区域,所述目标区域内包括所述二维点云;根据所述第一面积、所述第二面积和所述第三面积,确定所述第一可见边缘代价。
- 根据权利要求11所述的方法,其特征在于,所述根据所述第一面积、所述第二面积和所述第三面积,确定所述第一参考角度对应的所述第一可见边缘代价,包括:获取所述第一面积、所述第二面积和第一预设权重,获取第一预选可见边缘代价;根据所述第三面积和第二预设权重,获取第二预选可见边缘代价;根据所述第一预选可见边缘代价和所述第二预选可见边缘代价,确定所述第一可见边缘代价。
- 根据权利要求3~12任一项所述的方法,其特征在于,所述预设角度范围包括的角度为第一预设角度与第二预设角度之间的角度,所述第一预设角度减去所述第二预设角度等于90°,所述预设角度范围包括所述第一预设角度和/或所述第二预设角度。
- 根据权利要求1所述的方法,其特征在于,所述拟合特征包括如下中的至少一项:所述目标物体的拟合角度、所述目标物体的拟合方向、所述 目标物体拟合边界框、所述目标物体的拟合可见边;其中,所述拟合方向为从点云传感器坐标系的坐标原点出发旋转所述拟合角度后所对应的方向,所述点云传感器坐标系的坐标原点为所述点云传感器所在的位置。
- 根据权利要求1或14所述方法,其特征在于,所述根据所述拟合特征对所述目标物体进行拟合,包括:根据所述目标物体的点云,获取所述目标物体的拟合高度;根据所述拟合特征和所述拟合高度,对所述目标物体进行拟合。
- 根据权利要求15所述的方法,其特征在于,在所述拟合特征包括如下中的至少一项时:目标物体的拟合角度、所述目标物体的拟合方向、所述目标物体的拟合可见边,根据所述拟合特征和所述拟合高度,对所述目标物体进行拟合,包括:根据所述拟合特征,获取所述目标物体拟合边界框;根据所述拟合边界框和拟合高度,对所述目标物体进行拟合。
- 根据权利要求15所述的方法,其特征在于,根据所述目标物体的点云,获取所述目标物体的拟合高度,包括:获取所述目标物体的点云的拟合最大高度和拟合最小高度;根据所述拟合最大高度和拟合最小高度,获取所述拟合高度。
- 一种点云传感器,所述点云传感器用于搭载在移动平台上,其特征在于,包括:采集器,用于采集所述目标物体的点云;处理器,与所述采集器通信连接,用于执行如下操作:根据所述采集器采集的目标物体的点云,生成俯视图下的二维点云;确定所述二维点云中朝向所述移动平台的可见区域;根据所述可见区域,确定所述目标物体的拟合特征,并根据所述拟合特征对所述目标物体进行拟合。
- 根据权利要求18所述的点云传感器,其特征在于,所述处理器在用于执行根据所述可见区域,确定所述目标物体的拟合特征的操作时,具体用于:根据所述可见区域和所述二维点云,确定所述拟合特征。
- 根据权利要求19所述的点云传感器,其特征在于,所述拟合特征为所述目标物体的拟合角度,所述处理器在用于执行根据所述可见区域和所述二维点云,确定所述拟合特征的操作时,具体用于:根据所述可见区域、所述二维点云和预设角度范围,确定所述拟合角度。
- 根据权利要求20所述的点云传感器,其特征在于,所述处理器在用于执行根据所述可见区域、所述二维点云和预设角度范围,确定所述拟合角度的操作时,具体用于:根据所述可见区域、所述二维点云和预设角度范围,确定对应的可见边缘代价最小的所述拟合角度。
- 根据权利要求21所述的点云传感器,其特征在于,所述处理器在用于执行根据所述可见区域、所述二维点云和预设角度范围,确定对应的可见边缘代价最小的所述拟合角度的操作时,具体用于:从所述预设角度范围内选取至少一个参考角度;根据所述至少一个参考角度、所述可见区域和所述二维点云,确定对应的可见边缘代价最小的所述拟合角度。
- 根据权利要求22所述的点云传感器,其特征在于,在参考角度的数目为多个时,所述处理器在用于执行根据所述至少一个参考角度、所述可见区域和所述二维点云,确定对应的可见边缘代价最小的所述拟合角度的操作时,具体用于:对于多个参考角度中的任意一个第一参考角度,根据所述可见区域和所述二维点云,在所述第一参考角度的第一邻域范围内确定可见边缘代价最小的第一预选角度,所述第一邻域范围包括所述第一参考角度;确定各预选角度中对应的可见边缘代价最小的预选角度为所述拟合角度。
- 根据权利要求23所述的点云传感器,其特征在于,所述处理器在用于执行根据所述可见区域和所述二维点云,在所述第一参考角度的第一邻域范围内确定可见边缘代价最小的第一预选角度的操作时,具体用于:根据所述可见区域和所述二维点云,获取所述第一参考角度对应的第一可见边缘代价,并更新所述第一参考角度至所述第一邻域范围内的第一角度,根据所述可见区域和所述二维点云,获取所述第一角度对应的第二可见边缘代价;根据所述第二可见边缘代价和所述第一可见边缘代价之间的关系,更新所述第一角度至所述第一邻域范围内的第二角度;重复执行获取所述第一邻域范围内的角度对应的可见边缘代价、根据相邻两次获取到的可见边缘代价的关系在所述第一邻域范围内更新角度的操作,直至获取到可见边缘代价最小的所述第一预选角度。
- 根据权利要求22所述的点云传感器,其特征在于,在参考角度的数目为多个时,所述处理器在用于执行根据所述至少一个参考角度、所述可见区域和所述二维点云,确定对应的可见边缘代价最小的所述拟合角度的操作时,具体用于:根据所述可见区域和所述二维点云,在多个参考角度中确定对应的可见边缘代价最小的目标参考角度;根据所述可见区域和所述二维点云,在所述目标参考角度的第二邻域范围内确定对应的可见边缘代价最小的所述拟合角度,所述第二邻域范围包括所述目标参考角度。
- 根据权利要求25所述的点云传感器,其特征在于,所述处理器在用于执行根据所述可见区域和所述二维点云,在所述目标参考角度的第二邻域范围内确定可见边缘代价最小的所述拟合角度的操作时,具体用于:从所述目标参考角度的第二邻域范围内获取多个子参考角度;根据所述可见区域和所述二维点云,确定各子参考角度对应的可见边缘代价中最小的子参考角度为所述拟合角度。
- 根据权利要求24所述的点云传感器,其特征在于,所述处理器在用于执行根据所述可见区域和所述二维点云,获取所述第一参考角度对应的第一可见边缘代价的操作时,具体用于:根据所述二维点云在第一方向上的第一外包络矩形、所述可见区域和所述可见区域对应的第一边界和第二边界,获取所述第一可见边缘代价;其中,所述第一方向为从点云传感器坐标系的坐标原点出发旋转所述第一参考角度后所对应的方向,所述点云传感器坐标系的坐标原点为所述点云传感器所在的位置。
- 根据权利要求27所述的点云传感器,其特征在于,所述处理器在用于执行根据所述二维点云在第一方向上的第一外包络矩形、所述可见区域和 所述可见区域对应的第一边界和第二边界,获取所述第一可见边缘代价的操作时,具体用于:获取由所述可见区域、所述第一边界和第二边界限定的第一区域的第一面积;获取由所述第一外包络矩形朝向所述移动平台的可见边、所述第一边界和第二边界限定的第二区域的第二面积;获取所述第一外包络矩形位于目标区域外的边、所述第一边界和第二边界限定的第三区域的第三面积,所述目标区域为所述第一边界和第二边界之间的区域,所述目标区域内包括所述二维点云;根据所述第一面积、所述第二面积和所述第三面积,确定所述第一可见边缘代价。
- 根据权利要求28所述的点云传感器,其特征在于,所述处理器在用于执行根据所述第一面积、所述第二面积和所述第三面积,确定所述第一参考角度对应的所述第一可见边缘代价的操作时,具体用于:获取所述第一面积、所述第二面积和第一预设权重,获取第一预选可见边缘代价;根据所述第三面积和第二预设权重,获取第二预选可见边缘代价;根据所述第一预选可见边缘代价和所述第二预选可见边缘代价,确定所述第一可见边缘代价。
- 根据权利要求20~29任一项所述的点云传感器,其特征在于,所述预设角度范围包括的角度为第一预设角度与第二预设角度之间的角度,所述第一预设角度减去所述第二预设角度等于90°,所述预设角度范围包括所述第一预设角度和/或所述第二预设角度。
- 根据权利要求18所述的点云传感器,其特征在于,所述拟合特征包括如下中的至少一项:所述目标物体的拟合角度、所述目标物体的拟合方向、所述目标物体拟合边界框、所述目标物体的拟合可见边;其中,所述拟合方向为从点云传感器坐标系的坐标原点出发旋转所述拟合角度后所对应的方向,所述点云传感器坐标系的坐标原点为所述点云传感器所在的位置。
- 根据权利要求18或31所述的点云传感器,其特征在于,所述处理 器在用于执行根据所述拟合特征对所述目标物体进行拟合的操作时,具体用于:根据所述目标物体的点云,获取所述目标物体的拟合高度;根据所述拟合特征和所述拟合高度,对所述目标物体进行拟合。
- 根据权利要求32所述的点云传感器,其特征在于,在所述拟合特征包括如下中的至少一项时:所述目标物体的拟合角度、所述目标物体的拟合方向、所述目标物体的拟合可见边,所述处理器在用于执行根据所述拟合特征和所述拟合高度,对所述目标物体进行拟合的操作时,具体用于:根据所述拟合特征,获取所述目标物体拟合边界框;根据所述拟合边界框和拟合高度,对所述目标物体进行拟合。
- 根据权利要求32所述的点云传感器,其特征在于,所述处理器在用于执行根据所述目标物体的点云,获取所述目标物体的拟合高度的操作时,具体用于:获取所述目标物体的点云拟合最大高度和拟合最小高度;根据所述拟合最大高度和拟合最小高度,获取所述拟合高度。
- 一种移动平台,其特征在于,包括:点云传感器,用于采集所述目标物体的点云;其中,所述点云传感器搭载于所述移动平台;处理器,与所述点云传感器通信连接,用于执行如下操作:根据所述点云传感器采集的目标物体的点云,生成俯视图下的二维点云,其中所述点云传感器搭载于移动平台;确定所述二维点云中朝向所述移动平台的可见区域;根据所述可见区域,确定所述目标物体的拟合特征,并根据所述拟合特征对所述目标物体进行拟合。
- 根据权利要求35所述的移动平台,其特征在于,所述处理器在用于执行根据所述可见区域,确定所述目标物体的拟合特征的操作时,具体用于:根据所述可见区域和所述二维点云,确定所述拟合特征。
- 根据权利要求36所述的移动平台,其特征在于,所述拟合特征为所述目标物体的拟合角度,所述处理器在用于执行根据所述可见区域和所述二维点云,确定所述拟合特征的操作时,具体用于:根据所述可见区域、所述二维点云和预设角度范围,确定所述拟合角度。
- 根据权利要求37所述的移动平台,其特征在于,所述处理器在用于执行根据所述可见区域、所述二维点云和预设角度范围,确定所述拟合角度的操作时,具体用于:根据所述可见区域、所述二维点云和预设角度范围,确定对应的可见边缘代价最小的所述拟合角度。
- 根据权利要求38所述的移动平台,其特征在于,所述处理器在用于执行根据所述可见区域、所述二维点云和预设角度范围,确定对应的可见边缘代价最小的所述拟合角度的操作时,具体用于:从所述预设角度范围内选取至少一个参考角度;根据所述至少一个参考角度、所述可见区域和所述二维点云,确定对应的可见边缘代价最小的所述拟合角度。
- 根据权利要求39所述的移动平台,其特征在于,在参考角度的数目为多个时,所述处理器在用于执行根据所述至少一个参考角度、所述可见区域和所述二维点云,确定对应的可见边缘代价最小的所述拟合角度的操作时,具体用于:对于多个参考角度中的任意一个第一参考角度,根据所述可见区域和所述二维点云,在所述第一参考角度的第一邻域范围内确定可见边缘代价最小的第一预选角度,所述第一邻域范围包括所述第一参考角度;确定各预选角度中对应的可见边缘代价最小的预选角度为所述拟合角度。
- 根据权利要求40所述的移动平台,其特征在于,所述处理器在用于执行根据所述可见区域和所述二维点云,在所述第一参考角度的第一邻域范围内确定可见边缘代价最小的第一预选角度的操作时,具体用于:根据所述可见区域和所述二维点云,获取所述第一参考角度对应的第一可见边缘代价,并更新所述第一参考角度至所述第一邻域范围内的第一角度,根据所述可见区域和所述二维点云,获取所述第一角度对应的第二可见边缘代价;根据所述第二可见边缘代价和所述第一可见边缘代价之间的关系,更新所述第一角度至所述第一邻域范围内的第二角度;重复执行获取所述第一邻域范围内的角度对应的可见边缘代价、根据相邻两次获取到的可见边缘代价的关系在所述第一邻域范围内更新角度的操作, 直至获取到可见边缘代价最小的所述第一预选角度。
- 根据权利要求39所述的移动平台,其特征在于,在参考角度的数目为多个时,所述处理器在用于执行根据所述至少一个参考角度、所述可见区域和所述二维点云,确定对应的可见边缘代价最小的所述拟合角度的操作时,具体用于:根据所述可见区域和所述二维点云,在多个参考角度中确定对应的可见边缘代价最小的目标参考角度;根据所述可见区域和所述二维点云,在所述目标参考角度的第二邻域范围内确定对应的可见边缘代价最小的所述拟合角度,所述第二邻域范围包括所述目标参考角度。
- 根据权利要求42所述的移动平台,其特征在于,所述处理器在用于执行根据所述可见区域和所述二维点云,在所述目标参考角度的第二邻域范围内确定可见边缘代价最小的所述拟合角度的操作时,具体用于:从所述目标参考角度的第二邻域范围内获取多个子参考角度;根据所述可见区域和所述二维点云,确定各子参考角度对应的可见边缘代价中最小的子参考角度为所述拟合角度。
- 根据权利要求43所述的移动平台,其特征在于,所述处理器在用于执行根据所述可见区域和所述二维点云,获取所述第一参考角度对应的第一可见边缘代价的操作时,具体用于:根据所述二维点云在第一方向上的第一外包络矩形、所述可见区域和所述可见区域对应的第一边界和第二边界,获取所述第一可见边缘代价;其中,所述第一方向为从点云传感器坐标系的坐标原点出发旋转所述第一参考角度后所对应的方向,所述点云传感器坐标系的坐标原点为所述点云传感器所在的位置。
- 根据权利要求41所述的移动平台,其特征在于,所述处理器在用于执行根据所述二维点云在第一方向上的第一外包络矩形、所述可见区域和所述可见区域对应的第一边界和第二边界,获取所述第一可见边缘代价的操作时,具体用于:获取由所述可见区域、所述第一边界和第二边界限定的第一区域的第一面积;获取由所述第一外包络矩形朝向所述移动平台的可见边、所述第一边界和第二边界限定的第二区域的第二面积;获取所述第一外包络矩形位于目标区域外的边、所述第一边界和第二边界限定的第三区域的第三面积,所述目标区域为所述第一边界和第二边界之间的区域,所述目标区域内包括所述二维点云;根据所述第一面积、所述第二面积和所述第三面积,确定所述第一可见边缘代价。
- 根据权利要求45所述的移动平台,其特征在于,所述处理器在用于执行根据所述第一面积、所述第二面积和所述第三面积,确定所述第一参考角度对应的所述第一可见边缘代价的操作时,具体用于:获取所述第一面积、所述第二面积和第一预设权重,获取第一预选可见边缘代价;根据所述第三面积和第二预设权重,获取第二预选可见边缘代价;根据所述第一预选可见边缘代价和所述第二预选可见边缘代价,确定所述第一可见边缘代价。
- 根据权利要求37~46任一项所述的移动平台,其特征在于,所述预设角度范围包括的角度为第一预设角度与第二预设角度之间的角度,所述第一预设角度减去所述第二预设角度等于90°,所述预设角度范围包括所述第一预设角度和/或所述第二预设角度。
- 根据权利要求35所述的移动平台,其特征在于,所述拟合特征包括如下中的至少一项:述目标物体的拟合角度、所述目标物体的拟合方向、所述目标物体拟合边界框、所述目标物体的拟合可见边;其中,所述拟合方向为从点云传感器坐标系的坐标原点出发旋转所述拟合角度后所对应的方向,所述点云传感器坐标系的坐标原点为所述点云传感器所在的位置。
- 根据权利要求35或48所述的移动平台,其特征在于,所述处理器在用于执行根据所述拟合特征对所述目标物体进行拟合的操作时,具体用于:根据所述目标物体的点云,获取所述目标物体的拟合高度;根据所述拟合特征和所述拟合高度,对所述目标物体进行拟合。
- 根据权利要求49所述的移动平台,其特征在于,在所述拟合特征包 括如下中的至少一项时:所述目标物体的拟合角度、所述目标物体的拟合方向、所述目标物体的拟合可见边,所述处理器在用于执行根据所述拟合特征和所述拟合高度,对所述目标物体进行拟合的操作时,具体用于:根据所述拟合特征,获取所述目标物体拟合边界框;根据所述拟合边界框和拟合高度,对所述目标物体进行拟合。
- 根据权利要求49所述的移动平台,其特征在于,所述处理器在用于执行根据所述目标物体的点云,获取所述目标物体的拟合高度的操作时,具体用于:获取所述目标物体的点云拟合最大高度和拟合最小高度;根据所述拟合最大高度和拟合最小高度,获取所述拟合高度。
- 一种移动平台,其特征在于,所述移动平台上搭载有权利要求18~34任一项所述的点云传感器。
- 一种计算机可读存储介质,包括程序或指令,当所述程序或指令在计算机上运行时,权利要求1~17任一所述的方法被执行。
Priority Applications (2)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201980005593.1A CN111316289A (zh) | 2019-04-10 | 2019-04-10 | 目标物体的拟合方法、点云传感器和移动平台 |
PCT/CN2019/082119 WO2020206639A1 (zh) | 2019-04-10 | 2019-04-10 | 目标物体的拟合方法、点云传感器和移动平台 |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
PCT/CN2019/082119 WO2020206639A1 (zh) | 2019-04-10 | 2019-04-10 | 目标物体的拟合方法、点云传感器和移动平台 |
Publications (1)
Publication Number | Publication Date |
---|---|
WO2020206639A1 true WO2020206639A1 (zh) | 2020-10-15 |
Family
ID=71161148
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
PCT/CN2019/082119 WO2020206639A1 (zh) | 2019-04-10 | 2019-04-10 | 目标物体的拟合方法、点云传感器和移动平台 |
Country Status (2)
Country | Link |
---|---|
CN (1) | CN111316289A (zh) |
WO (1) | WO2020206639A1 (zh) |
Families Citing this family (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN113597568A (zh) * | 2020-10-12 | 2021-11-02 | 深圳市大疆创新科技有限公司 | 数据处理方法、控制设备及存储介质 |
Citations (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN105404844A (zh) * | 2014-09-12 | 2016-03-16 | 广州汽车集团股份有限公司 | 一种基于多线激光雷达的道路边界检测方法 |
EP3332218A1 (en) * | 2015-08-03 | 2018-06-13 | TomTom Global Content B.V. | Methods and systems for generating and using localisation reference data |
CN109061703A (zh) * | 2018-06-11 | 2018-12-21 | 百度在线网络技术(北京)有限公司 | 用于定位的方法、装置、设备和计算机可读存储介质 |
CN109271880A (zh) * | 2018-08-27 | 2019-01-25 | 深圳清创新科技有限公司 | 车辆检测方法、装置、计算机设备和存储介质 |
CN109446886A (zh) * | 2018-09-07 | 2019-03-08 | 百度在线网络技术(北京)有限公司 | 基于无人车的障碍物检测方法、装置、设备以及存储介质 |
Family Cites Families (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
DE102005000652A1 (de) * | 2005-01-04 | 2006-07-13 | Robert Bosch Gmbh | Verfahren für die Objekterfassung |
WO2017157967A1 (en) * | 2016-03-14 | 2017-09-21 | Imra Europe Sas | Processing method of a 3d point cloud |
CN108875804B (zh) * | 2018-05-31 | 2019-12-20 | 腾讯科技(深圳)有限公司 | 一种基于激光点云数据的数据处理方法和相关装置 |
-
2019
- 2019-04-10 CN CN201980005593.1A patent/CN111316289A/zh active Pending
- 2019-04-10 WO PCT/CN2019/082119 patent/WO2020206639A1/zh active Application Filing
Patent Citations (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN105404844A (zh) * | 2014-09-12 | 2016-03-16 | 广州汽车集团股份有限公司 | 一种基于多线激光雷达的道路边界检测方法 |
EP3332218A1 (en) * | 2015-08-03 | 2018-06-13 | TomTom Global Content B.V. | Methods and systems for generating and using localisation reference data |
CN109061703A (zh) * | 2018-06-11 | 2018-12-21 | 百度在线网络技术(北京)有限公司 | 用于定位的方法、装置、设备和计算机可读存储介质 |
CN109271880A (zh) * | 2018-08-27 | 2019-01-25 | 深圳清创新科技有限公司 | 车辆检测方法、装置、计算机设备和存储介质 |
CN109446886A (zh) * | 2018-09-07 | 2019-03-08 | 百度在线网络技术(北京)有限公司 | 基于无人车的障碍物检测方法、装置、设备以及存储介质 |
Non-Patent Citations (1)
Title |
---|
YE GANG: "Research on Multi-target Detection and Tracking Algorithm of Autonomous Vehicle Based on 3D Lidar in Urban Environment", MASTER THESIS, 31 August 2016 (2016-08-31), CN, pages 1 - 72, XP009523617 * |
Also Published As
Publication number | Publication date |
---|---|
CN111316289A (zh) | 2020-06-19 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
EP3620823B1 (en) | Method and device for detecting precision of internal parameter of laser radar | |
US8072450B2 (en) | System and method for measuring a three-dimensional object | |
WO2021016751A1 (zh) | 一种点云特征点提取方法、点云传感***及可移动平台 | |
CN110031824A (zh) | 激光雷达联合标定方法及装置 | |
KR101858902B1 (ko) | 컴포넌트를 활용한 점군 데이터의 객체 위치정보 추출 시스템 | |
CN107766405A (zh) | 自动车辆道路模型定义*** | |
CN106887020A (zh) | 一种基于LiDAR点云的道路纵横断面获取方法 | |
JP2010541051A (ja) | ユーザ選択可能な建物形状のオプションを備えた地球空間モデリング・システム及び関連した方法 | |
WO2020168685A1 (zh) | 一种三维扫描视点规划方法、装置及计算机可读存储介质 | |
CN103761739A (zh) | 一种基于半全局能量优化的影像配准方法 | |
CN105184854B (zh) | 针对地下空间扫描点云成果数据的快速建模方法 | |
CN112197773B (zh) | 基于平面信息的视觉和激光定位建图方法 | |
CN110503723B (zh) | 一种牙列缺损数字模型观测线的确定方法 | |
WO2020206639A1 (zh) | 目标物体的拟合方法、点云传感器和移动平台 | |
CN108563915B (zh) | 车辆数字化仿真测试模型构建***及方法、计算机程序 | |
CN114299242A (zh) | 高精地图中图像处理方法、装置、设备以及存储介质 | |
CN110796735B (zh) | Nurbs曲面有限元板壳网格划分方法及计算机实现*** | |
CN116958218A (zh) | 一种基于标定板角点对齐的点云与图像配准方法及设备 | |
CN104036096A (zh) | 斜面上凸起特征映射为制造特征体积的映射方法 | |
CN114820505A (zh) | 一种动态目标的非接触测量方法 | |
CN112950708A (zh) | 一种定位方法、定位装置及机器人 | |
CN107146286B (zh) | 基于影像边缘特征的三维模型自动调整方法 | |
CN117315183B (zh) | 一种基于激光雷达构建三维地图和作业分析的方法 | |
CN116754847B (zh) | 海面复合目标远区电磁散射强度估计方法和装置 | |
CN117115239B (zh) | 用于远区电磁散射强度估计的入射线交点获取方法 |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
121 | Ep: the epo has been informed by wipo that ep was designated in this application |
Ref document number: 19924327 Country of ref document: EP Kind code of ref document: A1 |
|
NENP | Non-entry into the national phase |
Ref country code: DE |
|
122 | Ep: pct application non-entry in european phase |
Ref document number: 19924327 Country of ref document: EP Kind code of ref document: A1 |