WO2021129483A1

WO2021129483A1 - Method for determining point cloud bounding box, and apparatus

Info

Publication number: WO2021129483A1
Application number: PCT/CN2020/136800
Authority: WO
Inventors: 李晨鸣; 罗磊; 彭学明
Original assignee: 华为技术有限公司
Priority date: 2019-12-25
Filing date: 2020-12-16
Publication date: 2021-07-01
Also published as: CN113034539A; CN113034539B

Abstract

A method for determining a point cloud bounding box, and an apparatus. Said method comprises: dividing a first plane occupied by a first shape into N regions according to first location information (S51); determining a bounding box using each edge of M edges to serve as a reference edge, altogether obtaining M bounding boxes (S52); determining N sub-loss values, corresponding to the N regions, for each bounding box of the M bounding boxes (S53); determining a loss value corresponding to each bounding box according to the N sub-loss values corresponding to each bounding box (S54); determining the bounding box corresponding to the smallest loss value as a point cloud bounding box (S55). Relative positions between a processing apparatus and an object are taken into consideration during computation of sub-loss values, and as such said bounding box can reduce the influence of self-occlusion on an observation result and improve the accuracy of the observation result for a target object.

Description

Method and device for determining bounding box of point cloud

Cross references to related applications

This application claims the priority of a Chinese patent application filed with the State Intellectual Property Office of China, the application number is 201911358284.0, and the application name is "A method and device for determining the bounding box of a point cloud" on December 25, 2019, and its entire contents Incorporated in this application by reference.

Technical field

This application relates to the field of detection technology, and in particular to a method and device for determining the bounding box of a point cloud.

Background technique

At present, lidar can complete the tracking of the target object. For example, the lidar can measure the target object to obtain the point cloud of the target object, and then obtain the envelope of the target object according to the point cloud of the target object, and determine the bounding box of the target object according to the envelope of the target object, thus completing The subsequent process of identifying or tracking the target object.

There are many ways to determine the bounding box of the target object based on the envelope of the target object. For example, the bounding box of the target object can be directly obtained from the point cloud, but this method needs to process each point included in the point cloud, and the process is more complicated. Alternatively, the convex polygon envelope of the target object may be obtained according to the point cloud, and then the bounding box of the target object may be obtained according to the convex polygon envelope. Currently, multiple bounding boxes of the target object can be obtained, and then one of the multiple bounding boxes is selected as the final bounding box of the target object. This method is relatively simple, but when selecting the final bounding box from multiple bounding boxes, there is no reasonable selection method, which leads to a large error in the selected bounding box.

Summary of the invention

The embodiments of the present application provide a method and device for determining the bounding box of a point cloud, which are used to determine a more reasonable bounding box, so as to improve the accuracy of the target object recognition result.

In a first aspect, a method for determining a bounding box of a point cloud is provided. The method includes: dividing a first plane where a first graphic is located into N regions according to first position information, where the first position information is determined according to the processing device The position of the point cloud and the position of the point cloud are determined, the first graphic is a two-dimensional graphic obtained by projecting the point cloud onto the first plane, the first graphic includes M edges, and the point cloud is The point data set obtained by the processing device measuring the target object, N is an integer greater than or equal to 2, and M is a positive integer; using each of the M edges as a reference edge, a bounding box is determined to obtain a total of M Bounding boxes; determining that each bounding box in the M bounding boxes corresponds to the N sub-loss values of the N regions, and the first sub-loss value of the N sub-loss values is used to measure the first sub-loss value The portion corresponding to the area corresponding to a sub-loss value in the free space occupied by each bounding box, and the free space occupied by each bounding box is determined according to each bounding box and the N areas ; Determine the loss value corresponding to each bounding box according to the N sub-loss values; determine the bounding box corresponding to the smallest loss value as the bounding box of the point cloud.

The method may be executed by a processing device, for example, a detection device or a detection device capable of supporting the detection device to implement the functions required by the method, such as a chip system. Exemplarily, the detection device is a radar, such as a lidar or other radars, then the processing device may be a radar, or may be a device provided in the radar that can support the functions required by the radar to implement the method, such as a chip system, Or other functional modules.

In the embodiment of the present application, the first position information may, for example, indicate the relative position between the processing device and the point cloud, so that according to the relative position between the processing device and the point cloud, the first graphic in the point cloud is located. A plane is divided into N areas, and the sub-loss value is calculated for each area, which is equivalent to considering the relative position between the processing device and the object when calculating the sub-loss value. The bounding box determined in this way can minimize the influence of self-occlusion phenomenon on the observation result, improve the accuracy of the observation result of the target object, and then improve the subsequent processing process of tracking or identifying the target object. accuracy. In addition, when calculating the loss value, the embodiment of the application calculates the free space measurement of the corresponding area occupied by the bounding box, which helps to reduce the free space occupied by the bounding box, which makes the determined bounding box more accurate. .

In an optional implementation manner, dividing the first plane where the first graphic is located into N regions according to the first position information includes:

Determine at least one auxiliary line according to the first position information;

The first plane is divided into the N regions according to the at least one auxiliary line.

For example, the area in the point cloud that is close to the processing device, or the area with higher density in the point cloud, generally does not have self-occlusion. For these areas of the point cloud, the position of the point cloud can generally indicate the target object In other words, the processing device has a high degree of accuracy in the observation of this part of the point cloud, so that the shape of the target object corresponding to this part of the point cloud obtained by the processing device is more accurate; Areas with a long distance between processing devices, or areas with low density in the point cloud, may have the problem of self-occlusion. For these areas of the point cloud, the position of the point cloud represents the shape of the target object and the target object There may be deviations between the actual shapes of the point cloud, that is to say, the observation accuracy of this part of the point cloud by the processing device is low. In view of this, the embodiment of the present application proposes that the first plane can be divided into N regions according to the reliability of the observation, or according to the accuracy of the observation, so that the N regions can be considered separately. The reliability of the observation of the point cloud by the processing device is related to the location of the processing device and the position of the point cloud, or to the density of the point cloud. Therefore, the confidence level corresponding to the area is also related to the location and point cloud of the processing device. It is related to the position or the density of the point cloud. Therefore, the processing device can determine at least one auxiliary line according to the first position information, so that the first plane can be divided into N regions through the at least one auxiliary line. This method of dividing regions takes into account the position of the processing device and the position of the point cloud, so that the reliability of different regions is different, so that N regions can be considered separately, which helps to improve the accuracy of the determined bounding box. degree.

In an alternative embodiment,

The method also includes:

Using each of the at least one auxiliary line as a reference edge, determine a bounding box, and determine a total of P bounding boxes;

Determining the bounding box corresponding to the smallest loss value as the bounding box of the point cloud includes:

Determine the bounding box corresponding to the smallest loss value of the M bounding boxes and the P bounding boxes as the bounding box of the point cloud.

When determining the bounding box, M bounding boxes can be determined only based on M edges, without considering auxiliary lines, which helps to reduce the amount of calculation. Alternatively, in addition to determining the bounding box based on each of the M edges, the bounding box may also be determined based on the auxiliary line. For example, each of the at least one auxiliary line may be used as a reference edge to determine the bounding box, and a total of P bounding boxes may be determined. M bounding boxes are determined based on M edges, and P bounding boxes are determined based on at least one auxiliary line. When selecting a bounding box, you can choose from M+P bounding boxes, which expands the selection range and facilitates more selection. Appropriate bounding box.

In an optional implementation manner, the N areas correspond to N confidence levels, and the N confidence levels are used to indicate the accuracy of the point data in the N areas to characterize the target object.

In the embodiment of the present application, the first plane may be divided into N regions according to the reliability of the observation, or according to the accuracy of the observation, so as to facilitate the consideration of the N regions. For example, the degree of trustworthiness, or the accuracy of observation, can be reflected by the degree of confidence. For an area of the first plane, if the processing device has a higher degree of confidence in observing the area, the confidence of the area is higher, and if the processing device has a lower degree of reliability in observing the area, then The lower the confidence in this area. Therefore, the confidence level of an area can indicate the accuracy of the point data in the area to characterize the target object. Then, N regions correspond to N confidence levels. Quantifying the degree of trustworthiness or the accuracy of observations with confidence makes the calculation process more convenient.

In an optional implementation manner, the N confidence levels are determined according to the density of the point cloud and/or the distance from the point cloud to the processing device.

In the perception process of the processing device, areas in the point cloud with different relative positions to the processing device, or areas with different densities in the point cloud, may have different levels of trustworthiness. For example, the area in the point cloud that is close to the processing device, or the area with higher density in the point cloud, generally does not have self-occlusion. For these areas of the point cloud, the position of the point cloud can generally indicate the target object The real shape of the point cloud, that is, the processing device has a high degree of reliability in the observation of this part of the point cloud, so that the shape of the target object corresponding to this part of the point cloud obtained by the processing device is more accurate; and for the point cloud Areas that are far away from the processing device, or areas with low density in the point cloud, may have the problem of self-occlusion. For these areas of the point cloud, the position of the point cloud represents the shape of the target object. There may be deviations between the actual shapes of the target object, that is, the processing device has a low degree of reliability in the observation of this part of the point cloud. Therefore, the trustworthiness of the area can be determined according to the density of the point cloud, or the trustworthiness of the area can be determined according to the distance from the point cloud to the processing device, or it can also be determined according to the density of the point cloud and the distance from the point cloud to the processing device The trustworthiness of the region, or the trustworthiness of the region can also be determined based on other factors.

In an optional embodiment, the first figure is convex polygonal or elliptical.

After the point cloud of the target object is obtained, the point cloud can be projected to the first plane, and a two-dimensional graph can be obtained after the projection, which is called, for example, the first graph. For example, the first figure may be a convex polygon, or it may be an ellipse or the like. This can also be understood as obtaining the convex polygonal envelope or the elliptical envelope of the point cloud based on the point cloud. Of course, the first graphic may also have other shapes, which are related to the shape of the point cloud projected on the first plane, and there is no specific limitation.

In an optional implementation manner, the M bounding boxes include a first bounding box, the first bounding box corresponds to the N sub-loss values, and the N sub-loss values include the first sub-loss Value, the first sub-loss value corresponds to the first area in the N areas, and the first sub-loss value is the first area in the free space occupied by the first bounding box The free space occupied by the first bounding box is determined according to the first bounding box and the N regions.

The relationship between bounding box, sub-loss value and area is introduced here. For each bounding box of the M bounding boxes, this relationship is similar, so here only the first bounding box is taken as an example. The first bounding box is, for example, any one of M bounding boxes. Or, if P bounding boxes are also determined according to at least one auxiliary line, then the first bounding box can be considered as any one of the M+P bounding boxes.

In an optional implementation manner, the loss value corresponding to the first bounding box satisfies the following relationship:

cost=a ₁ f(S ₁ )+a ₂ f(S ₂ )+……+a _N f(S _N )

Wherein, cost represents the loss value corresponding to the first bounding box, a ₁ , a ₂ ,..., a _N represents N coefficients, and the N coefficients correspond to the N regions one to one, S ₁ , S ₂ ,..., S _N represents the area of the corresponding N parts of the N regions in the free space occupied by the first bounding box, and f(x) represents a function of x.

The sub-loss value of a region is related to the area of the corresponding part of the region in the free space occupied by the bounding box. For example, one way is that the sub-loss value is the area, or the other way is that the sub-loss value is a function of the area . For example, the first plane is divided into N regions. For a bounding box, the areas of the corresponding N parts of the N regions in the free space occupied by the bounding box are, for example, S ₁ , S ₂ , ... ,S _N. Then for the bounding box, the N sub-loss values corresponding to the N regions are expressed as f(S ₁ ), f(S ₂ ),..., f(S _N ), where f(x) represents x The function. In the embodiment of the present application, f(x) may be a monotonically increasing function, such as a linear function, an exponential function, or a logarithmic function. There may be many ways to obtain the corresponding loss value according to the N sub-loss values, and the above relationship is just one example.

In an optional implementation manner, a _i is one of the N coefficients, the a _i is determined according to the i-th confidence level, and the i-th confidence level is one of the N confidence levels, And the a _i and the i-th confidence degree both correspond to the i-th region in the N regions, and 1≦i≦N.

Each of the N coefficients can be greater than or equal to zero. Among the N coefficients, some of the coefficients may be the same, or the N coefficients may all be different. For example, N coefficients can be determined according to the degree of confidence. A principle for determining N coefficients is, for example, when calculating the loss value, the sub-loss value corresponding to the observation accurate area should be as large as possible after the coefficient is processed, and the sub-loss value corresponding to the observing fuzzy area should be as large as the result after the coefficient processing It should be as small as possible. In other words, the sub-loss value corresponding to the area with higher confidence should be as large as possible according to the result of coefficient processing, and the sub-loss value corresponding to the area with lower confidence should be as small as possible according to the result of coefficient processing. For each bounding box of the M bounding boxes, the processing device can determine the corresponding N sub-loss values in a similar manner, and determine the loss value according to the N sub-loss values. In this way, M loss values corresponding to M bounding boxes can be determined. In the embodiment of the present application, the free space occupied by the bounding box is used as the object to be measured by the sub-loss value, which can reduce the influence of the self-occlusion phenomenon on the observation result, compared to the area in the first plane as the measurement object In terms of the method, the method of the embodiment of the present application is more reasonable.

In a second aspect, a processing device is provided, for example, the processing device is the aforementioned processing device. The processing device is used to execute the method in the foregoing first aspect or any possible implementation manner. Specifically, the processing device may include a module for executing the method in the first aspect or any possible implementation manner, for example, including a processing module and an acquisition module. Exemplarily, the processing device is a detection device, or a chip system or other components provided in the detection device. Exemplarily, the detection device is a radar. among them,

The processing module is configured to divide the first plane where the first graphic is located into N regions according to the first position information, the first position information is determined according to the position of the processing device and the position of the point cloud, the first A graph is a two-dimensional graph obtained by projecting the point cloud onto the first plane, the first graph includes M edges, and the point cloud is a point data set obtained by the acquisition module measuring a target object , N is an integer greater than or equal to 2, and M is a positive integer;

The processing module is further configured to use each of the M edges as a reference edge to determine a bounding box, and obtain a total of M bounding boxes;

The processing module is further configured to determine that each of the M bounding boxes corresponds to the N sub-loss values of the N regions, and the first sub-loss value of the N sub-loss values is used for Measure the corresponding part of the area corresponding to the first sub-loss value in the free space occupied by each bounding box, and the free space occupied by each bounding box is based on each bounding box and the N Determined by a region;

The processing module is further configured to determine the loss value corresponding to each bounding box according to the N sub-loss values;

The processing module is further configured to determine the bounding box corresponding to the smallest loss value as the bounding box of the point cloud.

In an optional implementation manner, the processing module is configured to divide the first plane where the first graphic is located into N regions according to the first position information in the following manner:

In an alternative embodiment,

The processing module is further configured to use each of the at least one auxiliary line as a reference edge to determine a bounding box, and determine P bounding boxes in total;

The processing module is configured to determine the bounding box corresponding to the smallest loss value as the bounding box of the point cloud in the following manner:

In an optional implementation manner, the N regions correspond to N confidence levels, and the N confidence levels are used to indicate the accuracy of the point data in the N regions to characterize the target object.

In an optional embodiment, the first figure is convex polygonal or elliptical.

cost=a ₁ f(S ₁ )+a ₂ f(S ₂ )+……+a _N f(S _N )

In a third aspect, a processing device is provided. The processing device is, for example, the aforementioned processing device. The processing device includes a processor and a communication interface. For example, the processor can implement the function of the processing module as described in the second aspect, and the communication interface can implement the function of the transceiver module as described in the second aspect. Optionally, the processing device may further include a memory for storing computer instructions. The processor, the communication interface, and the memory are coupled with each other, and are used to implement the methods described in the first aspect or various possible implementation manners. Alternatively, the processing device may not include a memory, and the memory may be located outside the processing device. For example, when the processor executes the computer instructions stored in the memory, the processing device is caused to execute the method in the foregoing first aspect or any one of the possible implementation manners. Exemplarily, the processing device is a detection device, or a chip system or other components provided in the detection device. Exemplarily, the detection device is a radar.

Wherein, if the processing device is a detection device, the communication interface is implemented, for example, by the transceiver (or transmitter and receiver) in the detection device. For example, the transceiver is realized by the antenna, feeder, and codec in the detection device.器, etc. to achieve. Or, if the processing device is a chip set in the detection device, the communication interface is, for example, the input/output interface of the chip, such as input/output pins, etc., and the communication interface is connected to the radio frequency transceiver component in the detection device to pass the radio frequency The transceiver component realizes the sending and receiving of information.

In an optional implementation manner, the radar device and the processing device jointly implement the methods provided in the first aspect or various optional implementation manners. The processing device is a processor provided outside the radar device, or may also be a processor provided in the radar device, such as a central processing unit. Specifically, the radar device is used to execute the content executed by the aforementioned detector or acquisition module, and the processing device is used to execute the content executed by the aforementioned processor or processing module. That is to say, the method provided in this application can be implemented by the radar device and the processing device. Realize together.

In a fourth aspect, a detection system is provided, which includes the processing device described in the second aspect or the processing device described in the third aspect.

In a fifth aspect, a smart car is provided, which includes the processing device described in the second aspect or the processing device described in the third aspect. Alternatively, the smart car is the processing device described in the second aspect or the processing device described in the third aspect.

In a sixth aspect, a computer-readable storage medium is provided, the computer-readable storage medium is used to store computer instructions, and when the computer instructions run on a computer, the computer executes the first aspect or any one of the above The methods described in the possible implementations.

In a seventh aspect, a chip is provided, the chip includes a processor and a communication interface, the processor is coupled with the communication interface, and is configured to implement the method provided in the first aspect or any of the optional implementation manners above .

Optionally, the chip may also include a memory. For example, the processor may read and execute a software program stored in the memory to implement the above-mentioned first aspect or any one of the optional implementation manners. method. Alternatively, the memory may not be included in the chip, but located outside the chip, which is equivalent to that the processor can read and execute the software program stored in the external memory to implement the first aspect or Any of the methods provided by the alternative implementations.

In an eighth aspect, a computer program product containing instructions is provided, the computer program product is used to store computer instructions, and when the computer instructions run on a computer, the computer executes the first aspect or any one of the above The methods described in the possible implementations.

In the embodiment of the present application, the relative position between the processing device and the object is considered when calculating the sub-loss value. The bounding box determined in this way can minimize the effect of self-occlusion on the observation result and improve The accuracy of the observation result of the target object can further improve the accuracy of subsequent processing processes such as tracking or recognizing the target object.

Description of the drawings

Figure 1 is a schematic diagram of the self-occlusion phenomenon of the determined bounding box;

Figure 2 is a flow chart of the basic processing procedure of the original point cloud by the lidar;

Figures 3A to 3C are schematic diagrams of three kinds of point cloud envelopes;

FIG. 4 is a schematic diagram of an application scenario of an embodiment of the application;

FIG. 5 is a flowchart of a method for determining the bounding box of a point cloud according to an embodiment of the application;

6A and 6B are two schematic diagrams of dividing areas in an embodiment of this application;

FIG. 6C is a schematic diagram of determining auxiliary lines in an embodiment of this application;

FIG. 7 is a schematic diagram of a target object in an embodiment of the application;

FIG. 8A is a schematic diagram of a convex polygon obtained according to a target object in an embodiment of the application; FIG.

FIG. 8B is a schematic diagram of the sub-loss value of the area in an embodiment of the application;

FIG. 9A is another schematic diagram of a convex polygon obtained according to a target object in an embodiment of the application; FIG.

FIG. 9B is another schematic diagram of the sub-loss value of the area in the embodiment of the application; FIG.

FIG. 10 is a schematic structural diagram of a processing device provided by an embodiment of this application;

FIG. 11 is a schematic structural diagram of a processing device provided by an embodiment of this application;

FIG. 12 is a schematic structural diagram of a processing device provided by an embodiment of this application;

FIG. 13 is a schematic structural diagram of a processing device provided by an embodiment of this application.

Detailed ways

In order to make the objectives, technical solutions, and advantages of the embodiments of the present application clearer, the embodiments of the present application will be further described in detail below with reference to the accompanying drawings.

Hereinafter, some terms in the embodiments of the present application will be explained to facilitate the understanding of those skilled in the art.

1) The processing device, or may also be called a detection device, such as a sensor, and the sensor is, for example, a radar, such as a light detection and ranging (lidar), or other types of radar. Alternatively, the sensor may also be a sensor installed on the radar and used to collect the point cloud of the target object.

2) Free space refers to the space occupied by non-objects. In the same way, invading free space means that there is no point cloud of an object in the space, but the space is considered to belong to an object.

3) Self-occlusion refers to a phenomenon in which a certain part of the object occludes another part of the object when the sensor is observing an object, causing the occluded part of the object to be invisible or the observation result of the occluded part is inaccurate.

4) In the embodiments of the present application, "at least one" refers to one or more, and "multiple" refers to two or more. "And/or" describes the association relationship of the associated objects, indicating that there can be three relationships, for example, A and/or B, which can mean: A alone exists, A and B exist at the same time, and B exists alone, where A, B can be singular or plural. The character "/" generally indicates that the associated objects before and after are in an "or" relationship. "The following at least one item (a)" or similar expressions refers to any combination of these items, including any combination of a single item (a) or a plurality of items (a). For example, at least one item (a) of a, b, or c can mean: a, b, c, ab, ac, bc, or abc, where a, b, and c can be single or multiple .

And, unless otherwise stated, the ordinal numbers such as "first" and "second" mentioned in the embodiments of this application are used to distinguish multiple objects, and are not used to limit the size, shape, content, and order of multiple objects. , Timing, priority or importance, etc. For example, the first bounding box and the second bounding box are only used to distinguish different bounding boxes, but do not indicate the difference in size, shape, priority, or importance of the two bounding boxes.

The foregoing introduces some terms and concepts involved in the embodiments of the present application, and the technical features involved in the embodiments of the present application are introduced below.

In the autonomous driving system/assisted driving system, it is necessary to solve the problems of perception, decision-making and execution. The so-called perception refers to perceiving the surrounding environment, such as perceiving objects in the surrounding environment, and recognizing and analyzing the perceived objects. Perception plays an extremely important role in the automatic driving system/assisted driving system. Lidar is one of the most important sensing sensors in automatic driving. The basic process from the original point cloud to the target perception can be referred to Figure 2. In Figure 2, the original point cloud is obtained by measuring the target object by lidar. After obtaining the original point cloud, the lidar can de-ground the original point cloud, that is, remove the points corresponding to the ground in the original point cloud, which is equivalent to eliminating ground interference. After that, the lidar performs clustering processing on the point cloud after ground removal. The obtained point cloud may include multiple objects. For example, the point cloud collected by the lidar includes objects such as pedestrians, vehicles, and houses. Then the lidar can cluster the points corresponding to each object together, which is equivalent to multiple categories of points obtained by clustering, which can respectively represent different objects. After the clustering process, the lidar can obtain the point cloud envelope corresponding to each object obtained after the clustering, or, after the clustering, multiple objects may be obtained, and only some of the objects may be lasers. For the objects that the radar wants to obtain, the lidar can also only obtain the point cloud envelope of these objects. After obtaining the point cloud envelope of the object, the lidar can determine the bounding box of the object based on the point cloud envelope, and then can track or recognize the corresponding object based on the bounding box.

Among them, the point cloud envelope is an effective simplified representation of the edge of the point cloud, which can provide great help for subsequent target tracking and other operations. In general, the main forms of point cloud envelopes are convex polygons (also known as convex polygonal envelopes), ellipses (also known as elliptical envelopes), and bounding boxes (also known as bounding boxes, also known as bounding boxes). Envelope for the bounding box) these three forms. It can be understood that the point cloud envelope may include several envelopes such as convex polygon envelope, ellipse envelope and bounding box envelope. Among them, the bounding box envelope can be understood as a type of convex polygonal envelope. For example, when the shape of the convex polygonal envelope is a rectangle, the convex polygonal envelope can also be regarded as a bounding box envelope. If the shape of the convex polygonal envelope is not a rectangle, the convex polygonal envelope is not a bounding box envelope. If this is the case, the bounding box envelope can be further obtained based on the convex polygon envelope. In the following, for simplicity, the bounding box envelope is also referred to as a bounding box, and the convex polygonal envelope is also referred to as a convex polygon.

The bounding box can be understood as an imaginary outer frame surrounding the object being detected. In digital image processing, the bounding box may be the coordinates of a rectangular frame that completely encloses the digital image when the digital image is placed on a page, canvas, screen, or other similar two-dimensional background. It can be simply understood that the bounding box is a geometric body with a slightly larger volume and simple characteristics to approximate a complex geometric object.

Refer to FIGS. 3A to 3C. FIG. 3A is a schematic diagram of a convex polygonal envelope, FIG. 3B is a schematic diagram of an elliptical envelope, and FIG. 3C is a schematic diagram of a bounding box envelope. In FIGS. 3A to 3C, the object is a vehicle as an example. In addition, you can refer to Table 1 for the advantages and disadvantages of the three types of envelopes.

Table 1

To	凸多边形Convex polygon	椭圆oval	边界框Bounding box
优点advantage	灵活、准确Flexible and accurate	稳定、简单Stable and simple	稳定、简单Stable and simple
不足insufficient	方法复杂，外形稳定性差Complex method, poor shape stability	准确性差Poor accuracy	准确性适中Moderate accuracy

Because the bounding box structure is stable, the parameters are simple, and the accuracy of the representation of most road traffic participation targets is good, the bounding box has become the mainstream solution for representing the point cloud envelope. In other words, after the point cloud envelope is obtained, the bounding box of the object can be obtained according to the point cloud envelope. It can be seen that one form of the point cloud envelope is the bounding box envelope. This form is equivalent to obtaining the point cloud envelope to obtain the bounding box. If the point cloud envelope is a convex polygon envelope or an ellipse envelope, after the point cloud envelope is obtained, the bounding box of the object can be further obtained according to the point cloud envelope.

For example, the bounding box can be obtained in two ways: Method one, obtain directly from the point cloud (that is, the obtained point cloud envelope is the bounding box envelope); Method two, use the two-dimensional convex polygon envelope of the point cloud, Get the final bounding box.

If you obtain the bounding box directly from the point cloud, you generally need to traverse all the points in the point cloud, and determine the final bounding box through some point-related loss function (such as area loss function or distance loss function, etc.). Therefore, this method has high computational complexity and the algorithm takes a long time.

However, the method of obtaining the bounding box through the convex polygon envelope generally requires less calculation and better real-time performance. However, since only the geometric structure of convex polygons can be used, the orientation stability of the bounding box is poor. For example, if the bounding box is determined based on the convex polygon envelope, multiple bounding boxes may be determined based on one convex polygon envelope, and one bounding box needs to be selected as the final bounding box. However, at present, there is no reasonable selection criterion when selecting the bounding box, and the selected bounding box is likely to be unreasonable.

For example, please refer to FIG. 1, where the solid rectangular box and the dotted rectangular box are two bounding boxes of the target object, and the two bounding boxes are obtained based on the convex polygon envelope of the point cloud, for example. The black dot in the rectangle represents the point cloud of the object. After determining the two bounding boxes, the lidar can select one of the two bounding boxes as the bounding box of the object. For example, in Figure 1, the bounding box finally selected by the lidar is the bounding box shown by the solid rectangular box. However, the sensor that collects the point cloud is located on the side of the solid rectangular frame, and a part of the solid rectangular frame is blocked between the point cloud and the sensor, causing the point cloud to be occluded. This is the so-called self-occlusion phenomenon. As a result, the observation result of the lidar on the target object is not reasonable enough, which will also affect the accuracy of the subsequent processing of tracking or identifying the target object.

In view of this, the technical solutions of the embodiments of the present application are provided. In the embodiment of the present application, the first position information may, for example, indicate the relative position between the processing device and the point cloud, so that according to the relative position between the processing device and the point cloud, the first graphic in the point cloud is located. A plane is divided into N areas, and the sub-loss value is calculated for each area, which is equivalent to considering the relative position between the processing device and the object when calculating the sub-loss value. The bounding box determined in this way can minimize the influence of self-occlusion on the observation results, and can improve the accuracy of subsequent processing procedures such as tracking or recognizing the target object, thereby improving the accuracy of the observation result of the target object. accuracy. In addition, when calculating the loss value, the embodiment of the application calculates the free space measurement of the corresponding area occupied by the bounding box, which helps to reduce the free space occupied by the bounding box, which makes the determined bounding box more accurate. .

Please refer to FIG. 4, which is a schematic diagram of an application scenario of an embodiment of this application. In Figure 4, there are two vehicles running on the same road. For example, a radar is installed on vehicle A. The radar can measure the surrounding environment to obtain the point cloud of objects in the surrounding environment, and can perform ground processing, clustering processing, and point cloud envelope processing on the obtained point cloud, so that the subsequent can be based on the obtained boundary The frame performs operations such as tracking or recognizing the corresponding object. For example, the radar can obtain the point cloud of vehicle B, and can perform ground processing, clustering, and point cloud envelope processing on the point cloud of vehicle B, so that vehicle B can be processed according to the obtained bounding box of vehicle B Operations such as tracking or identification. In FIG. 4, the position of the radar on the vehicle A is only an example, and the specific position of the radar on the vehicle is not limited to this.

Of course, FIG. 4 is only an example, and the application scenario of the embodiment of the present application is not limited to this. For example, the processing device provided by the embodiment of the present application may not be a radar but other equipment, and the processing device may not be installed on the vehicle, but on other equipment, such as a smart robot or a drone, or a processing device It can also be set individually.

The embodiment of the present application provides a method for determining the bounding box of a point cloud. Please refer to FIG. 5, which is a flowchart of the method. In the following introduction process, the application of this method to the network architecture shown in FIG. 4 is taken as an example.

For ease of introduction, in the following, the execution of the method by the processing device is taken as an example. Because this embodiment is applied to the scene shown in FIG. 4 as an example, the processing device described below may be a radar in the scene shown in FIG. 4, or a radar set in the scene shown in FIG. 4 Sensors in radar, etc.

S51. Divide the first plane where the first graphic is located into N areas according to the first position information, where N is an integer greater than or equal to 2.

The processing device is a device for collecting the point cloud, and the embodiment shown in FIG. 5 may be executed by the processing device. For example, when the processing device measures the target object, multiple points can be obtained. These points can be regarded as constituting a point data set (or called a point set). The point data set can correspond to the target object, and the point data set is also It is the point cloud of the target object.

After the point cloud of the target object is obtained, the point cloud can be projected to the first plane, and a two-dimensional graph can be obtained after the projection, which is called the first graph, for example. The first plane is, for example, a plane determined by the processing device and the target object. For example, if the target object is on the ground, the first plane may be a horizontal plane. For example, the horizontal plane may also be referred to as an XY plane. Alternatively, the first plane can also be a plane at any angle. For example, the target object is not on the ground, it may be floating in the air, or the target object is a vehicle, but the vehicle is driving on a slope, etc., then the first plane can be at any angle flat. In a more reasonable way, the first plane can be a plane with any angle other than the second plane. For example, the processing device and the target object are projected onto a plane, which can be regarded as a second plane.

For example, the first figure may be a convex polygon, or it may be an ellipse or the like. This can also be understood as obtaining the convex polygonal envelope or the elliptical envelope of the point cloud based on the point cloud. Of course, the first graphic may also have other shapes, which are related to the shape of the point cloud projected on the first plane, and there is no specific limitation.

The first figure includes, for example, M sides, and M is a positive integer. Wherein, if the first figure is a convex polygon, the M sides are the M sides of the convex polygon, and the M sides may include all sides of the convex polygon or part of the sides of the convex polygon. For example, if the convex polygon is a pentagon, M can be equal to 5 or less than 5. Because sometimes only part of the edges of the convex polygon can be used to obtain a more accurate result, the M edges do not need to include all the edges of the convex polygon, which also helps to reduce the amount of calculation. Or, if the first figure is an ellipse, the M edges are all or part of the edges of the ellipse used to be fitted. For example, by fitting 1132 edges, an ellipse can be obtained, and M can be equal to 1132 or less than 1132. The value of M may depend on the point cloud. If the shape of the point cloud is different, the value of M may also be different, so there is no restriction on the specific value of M.

The processing device may divide the first plane into N areas. Among them, during the sensing process of the processing device, areas in the point cloud with different relative positions from the processing device, or areas with different densities in the point cloud may have different observation accuracy. The so-called observation accuracy refers to the accuracy with which the point data in the area in the point cloud characterizes the target object. The higher the accuracy of the processing device's observation of an area in the point cloud, the higher the reliability of the area. On the contrary, the lower the accuracy of the processing device's observation of an area in the point cloud, indicating the reliability of the area The lower the degree. For example, the area in the point cloud that is close to the processing device, or the area with higher density in the point cloud, generally does not have self-occlusion. For these areas of the point cloud, the position of the point cloud can generally indicate the target object In other words, the processing device has a high degree of accuracy in the observation of this part of the point cloud, so that the shape of the target object corresponding to this part of the point cloud obtained by the processing device is more accurate; Areas with a long distance between processing devices, or areas with a low density in the point cloud, may have the problem of self-occlusion (for the introduction of self-occlusion phenomenon, please refer to the previous article). For these areas of the point cloud, the point cloud The position of indicates that there may be a deviation between the shape of the target object and the actual shape of the target object, that is, the observation accuracy of this part of the point cloud by the processing device is low.

In view of this, the embodiment of the present application proposes that the first plane can be divided into N regions according to the reliability of the observation, or according to the accuracy of the observation, so that the N regions can be considered separately. For example, the degree of trustworthiness, or the accuracy of observation, can be reflected by the degree of confidence. For an area of the first plane, if the processing device has a higher degree of confidence in observing the area, the confidence of the area is higher, and if the processing device has a lower degree of reliability in observing the area, then The lower the confidence in this area. Therefore, the confidence level of an area can indicate the accuracy of the point data in the area to characterize the target object. For example, an area with a higher degree of confidence can also be referred to as an accurate observation area, and an area with a lower degree of confidence can also be referred to as a fuzzy observation area.

According to the above introduction, the reliability of the observation of the point cloud by the processing device is related to the position of the processing device and the position of the point cloud, or the density of the point cloud. Therefore, the confidence level corresponding to the area is also related to the processing device. The position of is related to the position of the point cloud, or is related to the density of the point cloud. Generally speaking, the density of the point cloud is also related to the location of the processing device and the location of the point cloud. For example, an area in the point cloud that is closer to the processing device may have a higher density, while an area in the point cloud that is farther away from the processing device may have a lower density. Therefore, to characterize the observation accuracy of each area of the first plane, or to determine the confidence of each area of the first plane, the first position information can be used. The first position information can be based on the position of the processing device and the point cloud. For example, the first position information may reflect the relative position between the processing device and the point cloud. For example, an area in the first plane that is closer to the processing device may have a higher confidence level, while an area in the first plane that is farther away from the processing device may have a lower confidence level. . Therefore, the processing device can determine at least one auxiliary line according to the first position information, so that the first plane can be divided into N areas through the at least one auxiliary line. These N regions correspond to N confidence levels, where the regions and the confidence levels may have a one-to-one correspondence. Among the N confidences corresponding to the N regions, some of the confidences may be the same, or all of the N confidences are different. In addition, there is no restriction on the value of N. For example, the value of N can be larger, so that the granularity of the division is lower, and more accurate results can be obtained; or the value of N can also be smaller, which helps reduce the amount of calculation.

Among the divided N areas, each area may include point data in the point cloud, or some areas may not include point data in the point cloud. In other words, each of the N regions may have an intersection with the first graphic, or some regions may not have an intersection with the first graphic. The confidence level can be related to the density of the point cloud. It can be understood that the N regions can correspond to N confidence degrees, and the less point data included in the region, the lower the corresponding confidence level may be. The more point data of the region, the greater the confidence level may be.

In the embodiment of the present application, there is no limitation on the number of auxiliary lines, as long as the first plane can be divided into N regions by at least one auxiliary line. In addition, the essence of the auxiliary line is the dividing line of different confidence levels (or the dividing line of the regions corresponding to different confidence levels), so there is no restriction on the shape of the auxiliary line. For example, the auxiliary line can be any curve or polyline in the two-dimensional space, and if there are multiple auxiliary lines, the shapes of the different auxiliary lines can be the same, for example, all are curves or all straight lines, or they can be different, for example, The auxiliary line of is a curve, and the auxiliary line is a straight line.

For example, refer to FIG. 6A. In FIG. 6A, the first figure is a convex polygon and N=2 as an example, that is, the first plane is divided into two regions. Then, the first plane can be divided into 2 regions by the auxiliary line of curve 1 in FIG. 6A, or the first plane can be divided into 2 regions by the auxiliary line of straight line 2 in FIG. 6A. For these two areas, the distance to the processing device is different. The distance between the area 1 and the processing device is shorter, and the distance between the area 2 and the processing device is longer, and the corresponding confidence of the area 1 is The degree of confidence is higher, and the confidence degree corresponding to region 2 is lower. Area 1 can be referred to as the accurate observation area, and area 2 can be referred to as the observing fuzzy area.

For another example, refer to FIG. 6B. FIG. 6B also takes the first figure as a convex polygon and N=2 as an example, that is, the first plane is divided into two regions. Then, the first plane can be divided into 2 regions by the auxiliary line of curve 1 in FIG. 6B, or the first plane can be divided into 2 regions by the auxiliary line of straight line 2 in FIG. 6B. For these two areas, the distance to the processing device is different. The distance between the area 1 and the processing device is shorter, and the distance between the area 2 and the processing device is longer, and the corresponding confidence of the area 1 is The degree of confidence is higher, and the confidence degree corresponding to region 2 is lower. Area 1 can be referred to as the accurate observation area, and area 2 can be referred to as the observing fuzzy area. It can be seen from FIG. 6A and FIG. 6B that because the divided area is related to the relative position between the processing device and the point cloud, when the position of the processing device is different, the division result of the area will also be different.

Fig. 6A and Fig. 6B both take the example of N areas including point data. Or it is also possible that some of the N areas do not include point data. For example, in FIG. 6A, it is possible that area 1 includes point data and area 2 does not include point data; or in FIG. 6B, it is possible that area 1 includes point data, but area 2 does not include point data.

The processing device determines at least one auxiliary line according to the first position information. Preferably, the number of auxiliary lines may be one, or of course there may be multiple auxiliary lines. There may be multiple ways to determine the auxiliary line according to the first position information. Here, taking the first figure as a convex polygon, N=2, the number of auxiliary lines is 1, and the auxiliary line is a straight line as an example, a way to determine the auxiliary line is introduced: 1) The processing device determines to scan the target object When determining the leftmost scan line and the rightmost scan line, as shown by the two dashed lines in FIG. 6A or FIG. 6B; 2) the processing device selects the two scan lines and the two convex polygons (that is, the first pattern) The intersection points are two scanning endpoints, such as the two points shown in A and B in FIG. 6A or FIG. 6B; 3) the processing device determines the straight line 2 passing through these two endpoints as an auxiliary line.

As mentioned above, the auxiliary line can be a straight line or a curve. Taking the first figure as a convex polygon, N=2, the number of auxiliary lines being 1, and the auxiliary line being a curve as an example, we will introduce a way to determine the auxiliary line: 1) When the processing device determines to scan the target object The scan line at the far left and the scan line at the far right are shown by the two dashed lines in FIG. 6A or FIG. 6B. 2) The processing device divides ∠AOB into multiple parts. Refer to FIG. 6C, where the leftmost scan line and the rightmost scan line form an angle. For example, the angle formed by the leftmost scan line and the rightmost scan line is called ∠AOB. For example, the processing device can draw multiple rays from point O of ∠AOB, and divide the area covered by ∠AOB into multiple fan-shaped parts. For example, one part is the area covered by ∠AOC, and the other part is the area covered by ∠COD, where C and D represent two rays, and so on. For example, when the processing device scans, the scan line moves gradually, and may move a certain degree each time (for example, 0.2° each time). The processing device can divide the area covered by ∠AOB into multiple parts according to the scan line, for example The angle corresponding to each part is 0.2°. 3) For each part obtained by dividing ∠AOB, a critical point can be found. Connecting multiple critical points corresponding to these multiple parts can get a curve, which can be used as an auxiliary line. After connecting multiple critical points, you can also smooth the connected lines to obtain the final auxiliary line. For example, the final auxiliary line is curve 1 in FIG. 6C.

Take a part obtained by dividing ∠AOB as an example to introduce a way to obtain critical points. For example, for the area covered by ∠AOC, there may be an intersection with the first graph. Include one or more point data in the intersection. For example, the total number of point data included in the intersection is F. For example, the point data can be counted from any direction of the intersection (for example, from a direction close to the processing device to a direction away from the processing device) until reaching a certain position of the intersection. , The number of counted point data is Q, or the ratio of the number of counted point data to F is the first ratio, then the position can be determined as the position of the critical point. Wherein, the value of Q or the first ratio may be specified by an agreement, or may be determined by the processing device, or may also be pre-configured in the processing device.

In the above, the number of auxiliary lines is 1 as an example. Or, if the number of auxiliary lines is greater than 1, the processing device may also determine the auxiliary lines in other ways, and there is no restriction on the way of determining the auxiliary lines.

It can be seen that, under the premise of considering the relative positional relationship between the processing device and the point cloud, the different relative positional relationship between the processing device and the point cloud may cause different auxiliary lines to be obtained. In this way, the entire observation process of the processing device is taken into consideration.

S52: Using each of the M edges as a reference edge, determine a bounding box to obtain a total of M bounding boxes.

The first graph corresponds to M edges, and one edge is used as a reference edge to get a bounding box. Then, based on M edges, M bounding boxes can be obtained. In the embodiment of the present application, the shape of the bounding box is a rectangle as an example. For example, for one of the M edges, to determine the bounding box corresponding to this edge, you can use this edge as an edge of the bounding box, and the determined bounding box needs to include the first graphic, and Make the area of the bounding box except the first graphic as small as possible. According to this principle, the bounding box uniquely corresponding to this side can be determined.

Refer to FIG. 7, which is a schematic diagram of the target object, and the target object is, for example, a vehicle. The processing device, for example, measures the front part of the target object (the convex polygon in FIG. 7 represents the measurement of the front part) to obtain a point cloud corresponding to the front part. The processing device projects the point cloud onto the first plane to obtain a first graph. The first graph may refer to FIG. 8A, which is the convex polygon in FIG. 8A. The processing device determines at least one auxiliary line according to the first position information, and divides the first plane into N areas through the at least one auxiliary line. In FIG. 8A, the number of at least one auxiliary line is 1, and N=2 as an example. For example, for the first graphic in FIG. 8A, the M edges may include all the edges of the first graphic, or may also include part of the edges of the first graphic, for example, only include the auxiliary lines in the first graphic. The lower side.

For each of the M edges, the corresponding bounding box can be determined in the above manner, so that M bounding boxes can be obtained.

In addition, for M edges, if there are two edges that are parallel or vertical, then only a bounding box needs to be determined based on any one of the two edges, and it is not necessary to determine a bounding box based on both edges. Determine the bounding box. Because the bounding box determined based on such two edges may be the same, it is sufficient to determine it once, which helps to reduce the amount of calculation.

As an optional way, in order to obtain more bounding boxes and make the selection range wider, in addition to determining the bounding box based on each of the M sides, the bounding box may also be determined based on auxiliary lines. For example, each of the at least one auxiliary line may be used as a reference edge to determine the bounding box, and a total of P bounding boxes may be determined. Among them, the value of P can be the same as the number of at least one auxiliary line. For example, it is necessary to determine the bounding box according to the auxiliary line shown in FIG. 8A. Because the determined bounding box needs to include the first graphic, the auxiliary line cannot be used as an edge of the bounding box, but may be located in the bounding box.

The same is true if P bounding boxes are to be determined based on at least one auxiliary line. For M edges and at least one auxiliary line, if there are two edges that are parallel or perpendicular, you only need to determine a bounding box based on any one of the two edges. All edges determine the bounding box.

S53. Determine that each of the M bounding boxes corresponds to N sub-loss values of the N regions. In other words, for each bounding box of M bounding boxes, N sub-loss values can be determined. Of course, the sub-loss values corresponding to different bounding boxes may be the same or different.

For example, for a bounding box in M bounding boxes (the bounding box can be any one of the M bounding boxes), the bounding box corresponds to N sub-loss values, any one of the N sub-loss values , Can be used to measure the corresponding part of the region corresponding to the sub-loss value in the free space occupied by the bounding box. The area corresponding to the sub-loss value refers to the area corresponding to the sub-loss value among the N areas divided by the auxiliary line. For example, for each bounding box of M bounding boxes, the meaning of the sub-loss value is similar. The free space occupied by a bounding box can be determined according to the bounding box and N regions. For example, for the first bounding box, in the first plane where the first graphic is located, the auxiliary line is used as the boundary. For the area that has an intersection with the first graphic, the area located in the first bounding box and outside the first graphic The part can be called the corresponding part of the area in the free space occupied by the first bounding box; in addition, in the first plane where the first graph is located, with the auxiliary line as the boundary, for the area that does not intersect with the first graph, There is no corresponding part in the free space occupied by the first bounding box, or in other words, the area that does not intersect with the first graphic is empty in the free space occupied by the first bounding box, in other words, the first bounding box occupies In the free space of, the area that does not intersect with the first graph among the N areas is not included. It can be understood that the free space occupied by the first bounding box includes a portion of the N areas that overlaps with the first graph and is located in the first bounding box and outside the first graph. Of course, the free space occupied by the bounding box here corresponds to the bounding box. When the bounding boxes are different, the areas included in the free space occupied by the bounding boxes may be different. For each of the N regions of the bounding box, there are one-to-one corresponding sub-loss values, and there are a total of N sub-loss values. Each sub-loss value can be understood as representing the accuracy of the point cloud described by the bounding box in the corresponding region.

For a better understanding, the following description takes the first bounding box of M bounding boxes as an example. For the first bounding box, it can correspond to N sub-loss values. The N sub-loss values include, for example, the first sub-loss value. The first sub-loss value can be used to measure that the area corresponding to the first sub-loss value is in the first bounding box. The corresponding part of the occupied free space. The area corresponding to the first sub-loss value refers to the first area corresponding to the first sub-loss value among the N areas divided by the auxiliary line. In addition, the free space occupied by the first bounding box here is Refers to the free space occupied by the first bounding box, or in other words, the free space corresponding to the first bounding box. The free space occupied by the first bounding box has been introduced above.

In the embodiment of this application, for a bounding box, the sub-loss value of a region may be related to the area of the corresponding part of the region in the free space occupied by the region. The free space occupied refers to the bounding box. Free space occupied. For example, taking an edge of the convex polygon shown in FIG. 8A as a reference edge, a bounding box can be determined, and you can refer to FIG. 8B. In FIG. 8A, the plane where the convex polygon is located is divided into two regions by auxiliary lines, and both regions include point data. For the bounding box shown in FIG. 8B (the dashed box in FIG. 8B represents the bounding box), the oblique line represents the free space occupied by the bounding box. The first plane shown in FIG. 8B is divided into two regions, where the confidence level corresponding to region 1 is higher, and the confidence level corresponding to region 2 is lower. The corresponding part of area 1 in the free space occupied by the bounding box is the part shown by "/" in FIG. 8B, and the corresponding part of area 2 in the free space occupied by the bounding box is the part in FIG. 8B The part shown by "\". Then, the sub-loss value of area 1 may be related to the area of the part shown by "/" in FIG. 8B, and the sub-loss value of area 2 may be related to the area of the part shown by "\" in FIG. 8B.

For another example, taking one side of the convex polygon shown in FIG. 9A as a reference side, a bounding box can be determined, and you can refer to FIG. 9B. In FIG. 9A, the plane where the convex polygon is located is divided into two areas by the auxiliary line, but area 2 does not include point data, and area 1 includes point data. For the bounding box shown in FIG. 9B (the dashed box in FIG. 9B represents the bounding box), the oblique line represents the free space occupied by the bounding box. The first plane shown in FIG. 9B is divided into two regions, where the confidence level corresponding to region 1 is higher, and the confidence level corresponding to region 2 is lower. The corresponding part of area 1 in the free space occupied by the bounding box is the part marked by "/" in FIG. 9B, and area 2 has no corresponding part in the free space occupied by the bounding box. Then, the sub-loss value of area 1 may be related to the area of the part marked by "/" in FIG. 9B, and the sub-loss value of area 2 may be zero.

The sub-loss value of a region is related to the area of the corresponding part of the region in the free space occupied by the bounding box. For example, one implementation is that the sub-loss value is the area, or another implementation is that the sub-loss value is a function of the area. Among the N regions, for the regions that overlap with the first graph, because there are corresponding parts in the free space occupied by the bounding box, the sub-loss value corresponding to this type of region can be greater than 0; while in the N regions, for and The first graph has no intersection area, because there is no corresponding part in the free space occupied by the bounding box, so the sub-loss value corresponding to this type of area can be equal to zero.

For example, the first plane is divided into N regions. For a bounding box, the corresponding areas of the N regions in the free space occupied by the bounding box are, for example, S ₁ , S ₂ , ..., S _{N, respectively} . Then for the bounding box, the N sub-loss values corresponding to the N regions are expressed as f(S ₁ ), f(S ₂ ),..., f(S _N ), where f(x) represents x The function. In the embodiment of the present application, f(x) may be a monotonically increasing function, such as a linear function, an exponential function, or a logarithmic function.

S54: Determine the loss value corresponding to each bounding box according to the N sub-loss values corresponding to each bounding box.

For a bounding box of M bounding boxes, after obtaining the N sub-loss values corresponding to the bounding box, the processing device can obtain the loss value corresponding to the bounding box according to the N sub-loss values. There are many ways to obtain the loss value according to the sub-loss value, one of which is introduced below. For example, for the first bounding box of M bounding boxes, the sub-loss value corresponding to the first bounding box may satisfy the following relationship:

cost=a ₁ f(S ₁ )+a ₂ f(S ₂ )+……+a _N f(S _N ) (Formula 1)

In formula 1, cost represents the sub-loss value corresponding to the first bounding box, a ₁ , a ₂ ,..., a _N represents N coefficients, and these N coefficients correspond to N regions one-to-one. S ₁ , S ₂ ,..., S _N respectively represent the area of the corresponding N parts of the N regions in the free space occupied by the first bounding box. For example, S ₁ represents that the first region of the N regions is at the first boundary The area of the corresponding part of the free space occupied by the frame, S ₂ represents the area of the corresponding part of the second area of the N regions in the free space occupied by the first bounding box, and so on. f(S ₁ ), f(S ₂ ),..., f(S _N ) represents the N sub-loss values of N regions, for example, f(S ₁ ) represents the sub-loss value of the first region among the N regions, And so on.

Each of the N coefficients can be greater than or equal to zero. Among the N coefficients, some of the coefficients may be the same, or the N coefficients may all be different. For example, N coefficients can be determined according to the degree of confidence. A principle for determining N coefficients is, for example, when calculating the loss value, the weight of the sub-loss value corresponding to the accurate observation area should be as large as possible, and the weight of the sub-loss value corresponding to the observable fuzzy area should be as small as possible, or in other words, confidence The weight of the sub-loss value corresponding to the area with higher degree should be as large as possible, and the weight of the sub-loss value corresponding to the area with lower confidence degree should be as small as possible. For example, a _i is one of N coefficients, a _i can be determined according to the i-th confidence level, the i-th confidence level is one of the N confidence levels, and both a _i and the i-th confidence level correspond to N regions The i-th region in, 1≤i≤N.

For example, the coefficient a ₁ corresponds to area 1, the coefficient a ₂ corresponds to area 2, a ₁ can be determined according to the confidence of area 1, a ₂ can be determined according to the confidence of area 2, and the confidence of area 1 is greater than the confidence of area 2. Degree, then a ₁ can be greater than a ₂ .

For each bounding box of the M bounding boxes, the processing device can determine the corresponding N sub-loss values in a similar manner, and determine the loss value according to the N sub-loss values. In this way, M loss values corresponding to M bounding boxes can be determined. In the embodiment of this application, by taking the free space occupied by the bounding box as the object of the sub-loss value measurement, the effect of the self-occlusion phenomenon on the observation result can be minimized, as opposed to taking the area of the area in the first plane as the measurement result. In terms of objects, the method of the embodiment of this application is more reasonable.

In the embodiment of the present application, the processing device may first determine M bounding boxes, and after the M bounding boxes are all determined, respectively determine the N sub-loss values corresponding to each bounding box of the M bounding boxes; or, process The device may also determine the N sub-loss values corresponding to the bounding box after determining a bounding box, and then determine the next bounding box and its sub-loss values.

As an optional implementation manner, if P bounding boxes are determined according to at least one auxiliary line in S52, then for each bounding box of the P bounding boxes, the loss value can also be determined in a similar manner, then P loss values can be determined. If this is the case, it is equivalent to that the processing device has determined a total of M+P loss values.

In addition, if this is the case, the processing device may first determine M bounding boxes, and after the M bounding boxes are all determined, respectively determine the N sub-loss values corresponding to each bounding box of the M bounding boxes, and then determine P bounding boxes, after the P bounding boxes are determined, the N sub-loss values corresponding to each bounding box of the P bounding boxes are determined respectively; or, the processing device may first determine M+P bounding boxes, and wait for M After the +P bounding boxes are all determined, respectively determine the N sub-loss values corresponding to each bounding box in the M+P bounding boxes; alternatively, the processing device can also determine a bounding box in the M bounding boxes. Determine the N sub-loss values corresponding to the bounding box, and then determine the next bounding box and its sub-loss values. After all the sub-loss values corresponding to the M bounding boxes and M bounding boxes are determined, determine the P boundaries A bounding box in the frame, and then determine the N sub-loss values corresponding to the bounding box, and then determine the next bounding box and its corresponding sub-loss value in the P bounding boxes; or, the processing device may also determine a boundary After the box, the N sub-loss values corresponding to the bounding box are determined, and then the next bounding box and its sub-loss values are determined, without distinguishing whether the bounding box belongs to M bounding boxes or P bounding boxes.

S55. Determine the bounding box corresponding to the smallest loss value as the bounding box of the point cloud.

If the processing device only determines M bounding boxes based on the M edges, but does not determine the bounding boxes based on the auxiliary line, the processing device can determine the minimum of the M loss values corresponding to the M bounding boxes, and assign the minimum value to The bounding box is determined as the bounding box of the point cloud. In this way, there is no need to consider the bounding box corresponding to the auxiliary line, and the calculation amount can be reduced.

Or, if the processing device determines M bounding boxes based on M edges and P bounding boxes based on at least one auxiliary line, then the processing device can determine M loss values and P bounding boxes corresponding to the M bounding boxes The minimum value of the corresponding P loss values (M+P loss values in total), and the bounding box corresponding to the minimum value is determined as the bounding box of the point cloud. In this way, both the edge corresponding to the first graph and the auxiliary line can be considered. The range of options when selecting the bounding box of the point cloud is larger, which can make the selected bounding box more accurate.

For example, the processing device may determine the loss value corresponding to the bounding box after all the sub-loss values corresponding to the bounding box are determined; or the processing device may determine the boundary after determining the N sub-loss values corresponding to a bounding box The loss value corresponding to the box, and then the sub-loss value corresponding to the next bounding box and the corresponding loss value are determined.

In the embodiment of the present application, the first position information may, for example, indicate the relative position between the processing device and the point cloud, so that the first plane in the point cloud can be divided into N according to the relative position between the processing device and the point cloud. For each area, the sub-loss value is calculated separately for each area, which is equivalent to considering the relative position between the processing device and the object when calculating the sub-loss value. By determining the bounding box in this way, the influence of self-occlusion phenomenon on the observation results can be minimized, the stability and rationality of the bounding box are improved, the accuracy of the observation results of the target object can be improved, and the subsequent detection of the target object can be improved. The accuracy of the tracking or identification process. Moreover, when calculating the loss value in the embodiment of this application, it calculates the free space measurement of the corresponding area occupied by the bounding box, which also helps to reduce the influence of the self-occlusion phenomenon on the observation result, which also makes the determined bounding box More accurate.

The device used to implement the foregoing method in the embodiments of the present application will be described below in conjunction with the accompanying drawings. Therefore, all the above content can be used in the subsequent embodiments, and the repeated content will not be repeated.

The embodiment of the present application may divide the processing device into functional modules. For example, each functional module may be divided corresponding to each function, or two or more functions may be integrated into one functional module. The above-mentioned integrated modules can be implemented in the form of hardware or software functional modules. It should be noted that the division of modules in the embodiments of the present application is illustrative, and is only a logical function division, and there may be other division methods in actual implementation.

For example, in the case of dividing the functional modules of the processing apparatus in an integrated manner, FIG. 10 shows a possible schematic structural diagram of the processing apparatus involved in the foregoing embodiment of the present application. The processing device 10 is, for example, the processing device involved in the embodiment shown in FIG. 5, or the processing device 10 may also be a chip or other functional components provided in a detection device, and the detection device is, for example, a radar (or a radar device). The processing device 10 may include a processing module 1001 and a transceiver module 1002. When the processing device 10 is a radar, the processing module 1001 may be a processor, such as a baseband processor. The baseband processor may include one or more central processing units (CPU), and the transceiver module 1002 may be a transceiver, It can include antennas and radio frequency circuits. When the processing device 10 is a component with the aforementioned radar function, the processing module 1001 may be a processor, such as a baseband processor, and the transceiver module 1002 may be a radio frequency unit. When the processing device 10 is a chip system, the processing module 1001 may be a processor of the chip system and may include one or more central processing units, and the transceiver module 1002 may be an input and output interface of the chip system (for example, a baseband chip).

Among them, the processing module 1001 can be used to perform all operations performed by the processing device in the embodiment shown in FIG. 5 except for the operations of collecting point clouds, such as S51 to S55, and/or to support the operations described herein. Other processes of technology. The transceiver module 1002 may be used to perform all the collection operations performed by the processing device in the embodiment shown in FIG. 5, such as the operation of collecting the point cloud of the target object, and/or other processes used to support the technology described herein.

In addition, the transceiver module 1002 may be a functional module that can perform both sending and receiving operations. For example, the transceiver module 1002 may be used to perform all the sending and receiving operations performed by the processing device 10, for example, When performing a sending operation, the transceiver module 1002 can be considered as a sending module, and when performing a receiving operation, the transceiver module 1002 can be considered as a receiving module; or, the transceiver module 1002 can also be a collective term for two functional modules. They are a sending module and a receiving module. The sending module is used to complete the sending operation. For example, the sending module can be used to perform all the sending operations performed by the processing device 10, and the receiving module is used to complete the receiving operation. For example, the receiving module can be used to perform All receiving operations performed by the processing device 10.

Alternatively, the transceiver module 1002 may not belong to the processing device 10. For example, the transceiver module 1002 and the processing device 10 are both located in the same vehicle. The transceiver module 1002 is, for example, a communication unit in the vehicle. The transceiver module 1002 and the processing device 10 can communicate. At this time, the processing device 10 may not need to actively detect the target, and only perform processing based on the point cloud data received by the transceiver module 1002.

For example, the processing module 1001 is configured to divide the first plane where the first graphic is located into N regions according to the first position information, the first position information is determined according to the position of the processing device and the position of the point cloud, the The first graphic is a two-dimensional graphic obtained by projecting the point cloud onto the first plane, the first graphic includes M edges, and the point cloud is a set of point data obtained by the transceiver module 1002 measuring the target object , N is an integer greater than or equal to 2, and M is a positive integer;

The processing module 1001 is further configured to determine a bounding box using each of the M edges as a reference edge, and obtain a total of M bounding boxes;

The processing module 1001 is further configured to determine that each bounding box in the M bounding boxes corresponds to the N sub-loss values of the N regions, and the first sub-loss value of the N sub-loss values is used to measure The portion corresponding to the area corresponding to the first sub-loss value in the free space occupied by each bounding box, and the free space occupied by each bounding box is based on each bounding box and the N Regionally determined

The processing module 1001 is further configured to determine the loss value corresponding to each bounding box according to the N sub-loss values;

The processing module 1001 is further configured to determine the bounding box corresponding to the smallest loss value as the bounding box of the point cloud.

As an optional implementation manner, the processing module 1001 is configured to divide the first plane where the first graphic is located into N regions according to the first position information in the following manner:

As an optional implementation,

The processing module 1001 is further configured to use each of the at least one auxiliary line as a reference edge to determine a bounding box, and determine P bounding boxes in total;

The processing module 1001 is configured to determine the bounding box corresponding to the smallest loss value as the bounding box of the point cloud in the following manner:

As an optional implementation manner, the N areas correspond to N confidence levels, and the N confidence levels are used to indicate the accuracy of the point data in the N areas to characterize the target object.

As an optional implementation manner, the N confidence levels are determined according to the density of the point cloud and/or the distance from the point cloud to the processing device.

As an optional implementation manner, the first figure is convex polygonal or elliptical.

As an optional implementation manner, the M bounding boxes include a first bounding box, the first bounding box corresponds to the N sub-loss values, and the N sub-loss values include the first sub-loss values , The first sub-loss value corresponds to a first area in the N areas, and the first sub-loss value is a value corresponding to the first area in the free space occupied by the first bounding box Area, the free space occupied by the first bounding box is determined according to the first bounding box and the N areas.

As an optional implementation manner, the loss value corresponding to the first bounding box satisfies the following relationship:

cost=a ₁ f(S ₁ )+a ₂ f(S ₂ )+……+a _N f(S _N )

As an optional implementation manner, a _i is one of the N coefficients, the a _i is determined according to the i-th confidence level, the i-th confidence level is one of the N confidence levels, and The a _i and the i-th confidence degree both correspond to the i-th region in the N regions, and 1≦i≦N.

FIG. 11 is a schematic diagram of another possible structure of the processing apparatus provided by an embodiment of the application. The processing device 11 may include a processor 1101 and a transceiver 1102, the functions of which may correspond to the specific functions of the processing module 1001 and the transceiver module 1002 shown in FIG. 10 respectively, and will not be repeated here. Optionally, the processing device 11 may further include a memory 1103 for storing program instructions and/or data for the processor 1101 to read. Of course, the processing device 11 may not include the memory 1103, and the memory 1103 may be located outside the processing device 11.

Fig. 12 provides a schematic diagram of another possible structure of the processing device. The processing devices provided in FIGS. 10 to 12 can implement the functions of the processing devices in the foregoing embodiments. The processing devices provided in Figures 10 to 12 can be part or all of the radar device in the actual communication scenario, or can be a functional module integrated in the radar device or located outside the radar device, for example, can be a chip system, specifically to achieve the corresponding The function of the processing device shall prevail, and the structure and composition of the processing device shall not be specifically limited.

In this optional manner, the processing device 12 includes a transmitting antenna 1201, a receiving antenna 1202, and a processor 1203. Further optionally, the processing device 12 further includes a mixer 1204 and/or an oscillator 1205. Further optionally, the processing device 12 may also include a low-pass filter and/or a directional coupler, etc. Among them, the transmitting antenna 1201 and the receiving antenna 1202 are used to support the processing device 12 to perform radio communication, the transmitting antenna 1201 supports the transmission of radar signals, and the receiving antenna 1202 supports the reception of radar signals and/or the reception of reflected signals to finally realize the detection function. The processor 1203 performs some possible determination and/or processing functions. Further, the processor 1203 also controls the operation of the transmitting antenna 1201 and/or the receiving antenna 1202. Specifically, the signal to be transmitted is transmitted by the processor 1203 controlling the transmitting antenna 1201, and the signal received through the receiving antenna 1202 can be transmitted to the processor 1203 for corresponding processing. The various components included in the processing device 12 can be used to cooperate to execute the method provided in the embodiment shown in FIG. 5. Optionally, the processing device 12 may also include a memory for storing program instructions and/or data. Among them, the transmitting antenna 1201 and the receiving antenna 1202 may be set independently, or may be integratedly set as transmitting and receiving antennas to perform corresponding transmitting and receiving functions.

FIG. 13 is a schematic structural diagram of an apparatus 13 provided by an embodiment of this application. The device 13 shown in FIG. 13 may be the processing device itself, or may be a chip or circuit capable of completing the functions of the processing device, for example, the chip or circuit may be provided in a radar device. The apparatus 13 shown in FIG. 13 may include a processor 1301 (for example, the processing module 1001 may be implemented by the processor 1301, and the processor 1101 and the processor 1301 may be the same component, for example) and an interface circuit 1302 (for example, the transceiver module 1002 may be implemented by the interface circuit 1302, the transceiver 1102 and the interface circuit 1302 are, for example, the same component). The processor 1301 may enable the device 13 to implement the steps executed by the processing device in the method provided in the embodiment shown in FIG. 5. Optionally, the device 13 may further include a memory 1303, and the memory 1303 may be used to store instructions. The processor 1301 executes the instructions stored in the memory 1303 to enable the device 13 to implement the steps executed by the processing device in the method provided in the embodiment shown in FIG. 5.

Further, the processor 1301, the interface circuit 1302, and the memory 1303 can communicate with each other through internal connection paths, and transfer control and/or data signals. The memory 1303 is used to store a computer program, and the processor 1301 can call and run the computer program from the memory 1303 to control the interface circuit 1302 to receive or send a signal to complete the execution of the processing device in the method provided by the embodiment shown in FIG. 5 A step of. The memory 1303 may be integrated in the processor 1301, or may be provided separately from the processor 1301.

Optionally, if the device 13 is a device, the interface circuit 1302 may include a receiver and a transmitter. Wherein, the receiver and the transmitter may be the same component or different components. When the receiver and transmitter are the same component, the component can be called a transceiver.

Optionally, if the device 13 is a chip or a circuit, the interface circuit 1302 may include an input interface and an output interface, and the input interface and the output interface may be the same interface, or may be different interfaces respectively.

Optionally, if the device 13 is a chip or a circuit, the device 13 may not include the memory 1303, and the processor 1301 may read instructions (programs or codes) in the memory outside the chip or circuit to implement the implementation shown in FIG. 5. The steps in the method provided by the example are processed by the device.

Optionally, if the device 13 is a chip or a circuit, the device 13 may include a resistor, a capacitor, or other corresponding functional components, and the processor 1301 or the interface circuit 1302 may be implemented by corresponding functional components.

As an implementation manner, the function of the interface circuit 1302 may be implemented by a transceiver circuit or a dedicated chip for transceiver. The processor 1301 may be implemented by a dedicated processing chip, a processing circuit, a processor, or a general-purpose chip.

As another implementation manner, a general-purpose computer may be considered to implement the processing apparatus provided in the embodiments of the present application. That is, the program codes that realize the functions of the processor 1301 and the interface circuit 1302 are stored in the memory 1303, and the processor 1301 implements the functions of the processor 1301 and the interface circuit 1302 by executing the program codes stored in the memory 1303.

Among them, the functions and actions of the modules or units in the device 13 listed above are only exemplary descriptions, and the functional units in the device 13 can be used to execute the actions or processing procedures performed by the processing device in the embodiment shown in FIG. 5. In order to avoid repetitive descriptions, detailed descriptions are omitted here.

In yet another optional manner, when the processing device is implemented by software, it may be implemented in the form of a computer program product in whole or in part. The computer program product includes one or more computer instructions. When the computer program instructions are loaded and executed on the computer, all or part of the processes or functions described in the embodiments of the present application are realized. The computer may be a general-purpose computer, a special-purpose computer, a computer network, or other programmable devices. The computer instructions may be stored in a computer-readable storage medium, or transmitted from one computer-readable storage medium to another computer-readable storage medium. For example, the computer instructions may be transmitted from a website, computer, server, or data center. Transmission to another website, computer, server, or data center via wired (such as coaxial cable, optical fiber, digital subscriber line (DSL)) or wireless (such as infrared, wireless, microwave, etc.). The computer-readable storage medium may be any available medium that can be accessed by a computer or a data storage device such as a server or a data center integrated with one or more available media. The usable medium may be a magnetic medium, (for example, a floppy disk, a hard disk, and a magnetic tape), an optical medium (for example, a DVD), or a semiconductor medium (for example, a solid state disk (SSD)).

It should be noted that the processor included in the above-mentioned processing device for executing the method provided by the embodiment of the present application may be a central processing unit (CPU), a general-purpose processor, or a digital signal processor. DSP), application-specific integrated circuit (ASIC), field programmable gate array (FPGA) or other programmable logic devices, transistor logic devices, hardware components or any combination thereof. It can implement or execute various exemplary logical blocks, modules, and circuits described in conjunction with the disclosure of this application. The processor may also be a combination for realizing computing functions, for example, including a combination of one or more microprocessors, a combination of a DSP and a microprocessor, and so on.

The steps of the method or algorithm described in the embodiments of the present application may be implemented in a hardware manner, or may be implemented in a manner in which a processor executes software instructions. Software instructions can be composed of corresponding software modules, which can be stored in random access memory (RAM), flash memory, read-only memory (ROM), erasable programmable read-only Memory (erasable programmable read-only memory, EPROM), electrically erasable programmable read-only memory (EEPROM), register, hard disk, mobile hard disk, compact disc (read-only memory) , CD-ROM) or any other form of storage medium known in the art. An exemplary storage medium is coupled to the processor, so that the processor can read information from the storage medium and write information to the storage medium. Of course, the storage medium may also be an integral part of the processor. The processor and the storage medium may be located in the ASIC. In addition, the ASIC may be located in the processing device. Of course, the processor and the storage medium may also exist as discrete components in the processing device.

It is understandable that FIGS. 10 to 13 only show simplified designs of the processing device. In practical applications, the processing device may include any number of transceivers, processors, controllers, memories, and other possible components.

If the processing device does not include a transceiver module, an embodiment of the application also provides a detection system, which includes the processing device and communication unit that executes the processing device and the communication unit mentioned in the foregoing embodiment of the application, and the communication unit is used to execute the processing device in the foregoing processing device. The steps performed by the transceiver module (for example, the transceiver module 1002). The detection system may be a device, and each device is located in the device as a functional module of the device, or the detection system may also include multiple devices, and the processing device and the communication unit are located in different devices.

Through the description of the above embodiments, those skilled in the art can clearly understand that for the convenience and brevity of the description, only the division of the above-mentioned functional modules is used as an example for illustration. In practical applications, the above-mentioned functions can be allocated according to needs. It is completed by different functional modules, that is, the internal structure of the device is divided into different functional modules to complete all or part of the functions described above.

In the several embodiments provided in this application, it should be understood that the disclosed device and method may be implemented in other ways. For example, the device embodiments described above are merely illustrative. For example, the division of the modules or units is only a logical function division. In actual implementation, there may be other division methods, for example, multiple units or components may be divided. It can be combined or integrated into another device, or some features can be omitted or not implemented. In addition, the displayed or discussed mutual coupling or direct coupling or communication connection may be indirect coupling or communication connection through some interfaces, devices or units, and may be in electrical, mechanical or other forms.

The units described as separate parts may or may not be physically separate. The parts displayed as units may be one physical unit or multiple physical units, that is, they may be located in one place, or they may be distributed to multiple different places. . Some or all of the units may be selected according to actual needs to achieve the objectives of the solutions of the embodiments.

In addition, the functional units in the various embodiments of the present application may be integrated into one processing unit, or each unit may exist alone physically, or two or more units may be integrated into one unit. The above-mentioned integrated unit can be implemented in the form of hardware or software functional unit.

If the integrated unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a readable storage medium. Based on this understanding, the technical solutions of the embodiments of the present application are essentially or the part that contributes to the prior art, or all or part of the technical solutions can be embodied in the form of a software product, and the software product is stored in a storage medium. It includes several instructions to make a device (which may be a single-chip microcomputer, a chip, etc.) or a processor (processor) execute all or part of the steps of the methods described in the various embodiments of the present application. The aforementioned storage media include: U disk, mobile hard disk, ROM, RAM, magnetic disk or optical disk and other media that can store program codes.

The above are only specific implementations of the embodiments of this application, but the scope of protection of the embodiments of this application is not limited to this. Any changes or substitutions within the technical scope disclosed in this application shall be covered in the implementation of this application. Within the scope of protection of the case. Therefore, the protection scope of the embodiments of the present application should be subject to the protection scope of the claims.

Claims

A method for determining the bounding box of a point cloud is characterized in that it includes:

The first plane where the first graphic is located is divided into N regions according to the first position information, the first position information is determined according to the position of the processing device and the position of the point cloud, and the first graphic is A two-dimensional graph obtained by projecting a cloud onto the first plane, the first graph including M edges, the point cloud is a point data set obtained by the processing device measuring a target object, and N is greater than or equal to 2 Integer of, M is a positive integer;

Using each of the M edges as a reference edge, determine a bounding box to obtain a total of M bounding boxes;

It is determined that each bounding box in the M bounding boxes corresponds to the N sub-loss values of the N regions, and the first sub-loss value of the N sub-loss values is used to measure the first sub-loss value A corresponding area in the free space occupied by each bounding box, where the free space occupied by each bounding box is determined according to each bounding box and the N areas;

Determine the loss value corresponding to each bounding box according to the N sub-loss values;

The bounding box corresponding to the smallest loss value is determined as the bounding box of the point cloud.
The method according to claim 1, wherein the dividing the first plane where the first graphic is located into N regions according to the first position information comprises:

Determine at least one auxiliary line according to the first position information;

The first plane is divided into the N regions according to the at least one auxiliary line.
The method according to claim 2, wherein the method further comprises:

Using each of the at least one auxiliary line as a reference edge, determine a bounding box, and determine a total of P bounding boxes;

Determining the bounding box corresponding to the smallest loss value as the bounding box of the point cloud includes:

Determine the bounding box corresponding to the smallest loss value of the M bounding boxes and the P bounding boxes as the bounding box of the point cloud.
The method according to any one of claims 1 to 3, wherein the N areas correspond to N confidence levels, and the N confidence levels are used to indicate that the point data in the N areas represents a target The accuracy of the object.
The method according to claim 4, wherein the N confidence levels are determined according to the density of the point cloud and/or the distance from the point cloud to the processing device.
The method according to any one of claims 1 to 5, wherein the first figure is a convex polygon or an ellipse.
The method according to any one of claims 1 to 6, wherein the M bounding boxes comprise a first bounding box, the first bounding box corresponds to the N sub-loss values, and the N sub-losses The value includes the first sub-loss value, the first sub-loss value corresponds to the first area in the N areas, and the first sub-loss value is the free space occupied by the first bounding box Corresponding to the area of the first region, and the free space occupied by the first bounding box is determined according to the first bounding box and the N regions.
The method according to claim 7, wherein the loss value corresponding to the first bounding box satisfies the following relationship:

cost=a 1 f(S 1 )+a 2 f(S 2 )+……+a N f(S N )

Wherein, cost represents the loss value corresponding to the first bounding box, a 1 , a 2 ,..., a N represents N coefficients, and the N coefficients correspond to the N regions one to one, S 1 , S 2 ,..., S N represents the area of the corresponding N parts of the N regions in the free space occupied by the first bounding box, and f(x) represents a function of x.
The method according to claim 8, wherein a i is one of the N coefficients, the a i is determined according to the i-th confidence level, and the i-th confidence level is among the N confidence levels And the a i and the i-th confidence degree both correspond to the i-th region in the N regions, and 1≤i≤N.
A processing device, characterized in that it comprises a processing module and a transceiver module, wherein:

The processing module is configured to divide the first plane where the first graphic is located into N regions according to the first position information, the first position information is determined according to the position of the processing device and the position of the point cloud, the first A graph is a two-dimensional graph obtained by projecting the point cloud onto the first plane, the first graph includes M edges, and the point cloud is a set of point data obtained by measuring the target object by the transceiver module , N is an integer greater than or equal to 2, and M is a positive integer;

The processing module is further configured to use each of the M edges as a reference edge to determine a bounding box, and obtain a total of M bounding boxes;

The processing module is further configured to determine that each of the M bounding boxes corresponds to the N sub-loss values of the N regions, and the first sub-loss value of the N sub-loss values is used for Measure the corresponding part of the area corresponding to the first sub-loss value in the free space occupied by each bounding box, and the free space occupied by each bounding box is based on each bounding box and the N Determined by a region;

The processing module is further configured to determine the loss value corresponding to each bounding box according to the N sub-loss values;

The processing module is further configured to determine the bounding box corresponding to the smallest loss value as the bounding box of the point cloud.
The processing device according to claim 10, wherein the processing module is configured to divide the first plane on which the first graphic is located into N regions according to the first position information in the following manner:

Determine at least one auxiliary line according to the first position information;

The first plane is divided into the N regions according to the at least one auxiliary line.
The processing device according to claim 11, wherein:

The processing module is further configured to use each of the at least one auxiliary line as a reference edge to determine a bounding box, and determine P bounding boxes in total;

The processing module is configured to determine the bounding box corresponding to the smallest loss value as the bounding box of the point cloud in the following manner:

Determine the bounding box corresponding to the smallest loss value of the M bounding boxes and the P bounding boxes as the bounding box of the point cloud.
The processing device according to any one of claims 10 to 12, wherein the N areas correspond to N confidence levels, and the N confidence levels are used to indicate point data representations in the N areas The accuracy of the target object.
The processing device according to claim 13, wherein the N confidence levels are determined according to the density of the point cloud and/or the distance from the point cloud to the processing device.
The processing device according to any one of claims 10 to 14, wherein the first figure is a convex polygon or an ellipse.
The processing device according to any one of claims 10 to 15, wherein the M bounding boxes comprise a first bounding box, the first bounding box corresponds to the N sub-loss values, and the N sub-loss values The loss value includes the first sub-loss value, the first sub-loss value corresponds to the first area in the N areas, and the first sub-loss value is the free area occupied by the first bounding box. The area in the space corresponding to the first region, and the free space occupied by the first bounding box is determined according to the first bounding box and the N regions.
The processing device according to claim 16, wherein the loss value corresponding to the first bounding box satisfies the following relationship:

cost=a 1 f(S 1 )+a 2 f(S 2 )+……+a N f(S N )

Wherein, cost represents the loss value corresponding to the first bounding box, a 1 , a 2 ,..., a N represents N coefficients, and the N coefficients correspond to the N regions one to one, S 1 , S 2 ,..., S N represents the area of the corresponding N parts of the N regions in the free space occupied by the first bounding box, and f(x) represents a function of x.
The processing device according to claim 17, wherein a i is one of the N coefficients, the a i is determined according to the i-th confidence level, and the i-th confidence level is N confidence levels And the a i and the i-th confidence level both correspond to the i-th region in the N regions, and 1≦i≦N.
A computer-readable storage medium, wherein the computer-readable storage medium stores a computer program, and when the computer program runs on a computer, the computer executes any one of claims 1-9 The method described.
A chip, characterized by comprising a processor and a communication interface, the processor being used to read instructions to execute the method according to any one of claims 1-9.