CN117493353A

CN117493353A - Grid map updating method

Info

Publication number: CN117493353A
Application number: CN202311244782.9A
Authority: CN
Inventors: 徐建云; 苗振伟; 刘旻哲; 李坤; 卿泉
Original assignee: Wuzhou Online E Commerce Beijing Co ltd
Current assignee: Zhejiang Cainiao Chuancheng Network Technology Co ltd
Priority date: 2023-09-25
Filing date: 2023-09-25
Publication date: 2024-02-02

Abstract

The embodiment of the specification provides a raster image update method, which comprises the following steps: acquiring a reference grid graph containing a target object and an object of interest; determining a target area corresponding to the object of interest in the reference grid graph, and constructing a target distribution heat graph corresponding to the reference grid graph according to attribute information of the target area; determining the position information of the target object in the reference grid graph, and constructing an object route graph corresponding to the reference grid graph according to the position information; and constructing a grid distribution curve based on the target distribution heat map and the object road map, and updating the reference grid map into a target grid map according to the grid distribution curve.

Description

Grid map updating method

Technical Field

The embodiment of the specification relates to the technical field of automatic driving, in particular to a grid map updating method.

Background

With the vigorous development of the automotive industry, the demand for unmanned driving by the user population is growing. And unmanned technology is increasingly in demand for use in traffic, transportation and unmanned distribution scenarios. In the prior art, in order to realize unmanned driving and improve driving safety, collected point cloud data is usually mapped into a raster pattern by adopting point cloud rasterization processing, so that automatic driving is performed by using the raster pattern mapped with the point cloud data. The three-dimensional point cloud rasterization is mostly implemented by adopting a uniform rasterization mode under a Cartesian system. Although the uniform rasterization method can support unmanned driving, considering the complexity of driving scenes, if the grids are uniformly divided in a wide space, there is a problem in that both resolution and computational resources cannot be fully achieved. Namely, if the resolution is reduced to maintain real-time performance, the aim of saving the computing resources can be achieved, but the accuracy is lower; or to increase resolution to maintain performance, but to greatly increase the consumption of computing resources. There is therefore a need for an effective solution to the above problems.

Disclosure of Invention

In view of this, the present embodiment provides a raster image update method. One or more embodiments of the present specification relate to a grid map updating apparatus, a model training method, a model training apparatus, a computing device, a computer-readable storage medium, and a computer program that solve the technical drawbacks of the prior art.

According to a first aspect of embodiments of the present disclosure, there is provided a method for updating a raster image, including:

acquiring a reference grid graph containing a target object and an object of interest;

determining a target area corresponding to the object of interest in the reference grid graph, and constructing a target distribution heat graph corresponding to the reference grid graph according to attribute information of the target area;

determining the position information of the target object in the reference grid graph, and constructing an object route graph corresponding to the reference grid graph according to the position information;

and constructing a grid distribution curve based on the target distribution heat map and the object road map, and updating the reference grid map into a target grid map according to the grid distribution curve.

According to a second aspect of embodiments of the present disclosure, there is provided another method for updating a raster image, applied to a vehicle control end, including:

Acquiring a reference grid pattern comprising a target vehicle and a concerned vehicle;

determining a target area corresponding to the concerned vehicle in the reference grid graph, and constructing a target distribution heat graph corresponding to the reference grid graph according to attribute information of the target area;

determining the position information of the target vehicle in the reference grid map, and constructing an object route map corresponding to the reference grid map according to the position information;

and constructing a grid distribution curve based on the target distribution heat map and the object road map, and updating the reference grid map into a target grid map according to the grid distribution curve, wherein the target grid map is used for automatic driving of the target vehicle.

According to a third aspect of embodiments of the present specification, there is provided a model training method, comprising:

acquiring a reference grid graph containing a target object and an object of interest, and inputting the reference grid graph into an initial grid division model for processing to obtain predicted grid size information;

constructing a grid distribution curve based on the target distribution heat map and the object roadmap, and determining reference grid size information according to the grid distribution curve;

and calculating a loss value based on the predicted grid size information and the reference grid size information, and performing parameter adjustment on the initial grid division model by using the loss value until a target grid division model meeting the training stop condition is obtained.

According to a fourth aspect of embodiments of the present specification, there is provided a raster image update apparatus, including:

an acquisition module configured to acquire a reference raster pattern including a target object and an object of interest;

the first construction module is configured to determine a target area corresponding to the object of interest in the reference grid graph, and construct a target distribution heat graph corresponding to the reference grid graph according to attribute information of the target area;

a second construction module configured to determine position information of the target object in the reference raster pattern, and construct an object route map corresponding to the reference raster pattern according to the position information;

And the updating module is configured to construct a grid distribution curve based on the target distribution heat map and the object roadmap, and update the reference grid map into a target grid map according to the grid distribution curve.

According to a fifth aspect of embodiments of the present disclosure, there is provided another raster image update apparatus applied to a vehicle control terminal, including:

an acquisition raster pattern module configured to acquire a reference raster pattern including a target vehicle and a vehicle of interest;

the heat map constructing module is configured to determine a target area corresponding to the concerned vehicle in the reference grid map, and construct a target distribution heat map corresponding to the reference grid map according to attribute information of the target area;

a construction roadmap module configured to determine position information of the target vehicle in the reference raster pattern and construct an object roadmap corresponding to the reference raster pattern according to the position information;

and an updating grid graph module configured to construct a grid distribution curve based on the target distribution heat graph and the object roadmap, and update the reference grid graph to a target grid graph according to the grid distribution curve, wherein the target grid graph is used for automatic driving of the target vehicle.

According to a sixth aspect of embodiments of the present specification, there is provided a model training apparatus comprising:

the model processing module is configured to acquire a reference raster pattern comprising a target object and an object of interest, input the reference raster pattern into an initial raster division model for processing, and acquire predicted raster size information;

the distribution heat map constructing module is configured to determine a target area corresponding to the object of interest in the reference grid map and construct a target distribution heat map corresponding to the reference grid map according to attribute information of the target area;

a build object roadmap module configured to determine position information of the target object in the reference raster pattern and build an object roadmap corresponding to the reference raster pattern according to the position information;

a dimension information determining module configured to construct a grid distribution curve based on the target distribution heat map and the object roadmap, and determine reference grid dimension information from the grid distribution curve;

and the model parameter adjusting module is configured to calculate a loss value based on the predicted grid size information and the reference grid size information, and adjust parameters of the initial grid division model by utilizing the loss value until a target grid division model meeting training stop conditions is obtained.

According to a seventh aspect of embodiments of the present specification, there is provided a computing device comprising:

a memory and a processor;

the memory is configured to store computer-executable instructions that, when executed by the processor, implement the steps of the grid map updating method or the model training method described above.

According to an eighth aspect of embodiments of the present specification, there is provided a computer readable storage medium storing computer executable instructions which, when executed by a processor, implement the steps of the above-described grid map updating method or model training method.

According to a ninth aspect of embodiments of the present specification, there is provided a computer program, wherein the computer program, when executed in a computer, causes the computer to perform the steps of the above-described grid map updating method or model training method.

In order to achieve the goal of concentrating limited resources into a region of interest and improving the target detection effect without increasing time consumption, the grid map updating method provided by the embodiment can firstly obtain a reference grid map containing target objects and corresponding attention, then determine a target region corresponding to the attention objects in the reference grid map, so as to construct a target distribution heat map corresponding to the reference grid map according to attribute information of the target region, and embody the region of interest in the reference grid map through the heat map; meanwhile, position information of a target object is determined in the reference grid graph, so that an object route graph corresponding to the reference grid graph is constructed according to the position information, and a route area concerned in the reference grid graph is embodied through the route graph; at this time, considering that the area affecting the traveling of the target object is the area of interest and the route area, the target distribution heat map and the target route map can be used as the basis for dividing the reference grid map, so that when grid updating is performed, denser grids can be added for the area of interest and the route area, and sparser grids can be added for other areas; that is, a grid distribution curve is constructed based on the target distribution heat map and the target road map, and the attention in each direction is expressed by the curve, so that the reference grid map is updated to the target grid map. When the grid division is carried out, the focused content can be selected according to the running condition at the current moment to carry out dense grid division processing, and other areas adopt sparse grid division processing, so that the dynamic grid is realized, and the aims of balancing calculation resources and detecting accuracy are achieved.

Drawings

FIG. 1 is a schematic diagram of a raster image update method provided in one embodiment of the present disclosure;

FIG. 2 is a flow chart of a method for grid map update provided in one embodiment of the present disclosure;

FIG. 3 is a schematic diagram of a raster pattern in a raster pattern update method according to an embodiment of the present disclosure;

FIG. 4 is a schematic diagram of a raster pattern in another raster pattern update method provided in one embodiment of the present disclosure;

FIG. 5 is a process flow diagram of a raster image update method provided by one embodiment of the present disclosure;

FIG. 6 is a flow chart of another method for grid map update provided by one embodiment of the present disclosure;

FIG. 7 is a flow chart of a model training method provided in one embodiment of the present disclosure;

FIG. 8 is a schematic diagram of a device for updating a grid map according to an embodiment of the present disclosure;

FIG. 9 is a schematic diagram of another device for updating a raster image according to one embodiment of the present disclosure;

FIG. 10 is a schematic structural view of a model training device according to an embodiment of the present disclosure;

FIG. 11 is a block diagram of a computing device provided in one embodiment of the present description.

Detailed Description

In the following description, numerous specific details are set forth in order to provide a thorough understanding of the present description. This description may be embodied in many other forms than described herein and similarly generalized by those skilled in the art to whom this disclosure pertains without departing from the spirit of the disclosure and, therefore, this disclosure is not limited by the specific implementations disclosed below.

The terminology used in the one or more embodiments of the specification is for the purpose of describing particular embodiments only and is not intended to be limiting of the one or more embodiments of the specification. As used in this specification, one or more embodiments and the appended claims, the singular forms "a," "an," and "the" are intended to include the plural forms as well, unless the context clearly indicates otherwise. It should also be understood that the term "and/or" as used in one or more embodiments of the present specification refers to and encompasses any or all possible combinations of one or more of the associated listed items.

It should be understood that, although the terms first, second, etc. may be used in one or more embodiments of this specification to describe various information, these information should not be limited by these terms. These terms are only used to distinguish one type of information from another. For example, a first may also be referred to as a second, and similarly, a second may also be referred to as a first, without departing from the scope of one or more embodiments of the present description. The word "if" as used herein may be interpreted as "at … …" or "at … …" or "responsive to a determination", depending on the context.

Furthermore, it should be noted that, user information (including, but not limited to, user equipment information, user personal information, etc.) and data (including, but not limited to, data for analysis, stored data, presented data, etc.) according to one or more embodiments of the present disclosure are information and data authorized by a user or sufficiently authorized by each party, and the collection, use, and processing of relevant data is required to comply with relevant laws and regulations and standards of relevant countries and regions, and is provided with corresponding operation entries for the user to select authorization or denial.

First, terms related to one or more embodiments of the present specification will be explained.

And (3) point cloud: and a three-dimensional point data set of the actual road surface is acquired through a three-dimensional laser scanning instrument.

A grid: in image editing, a grid is a pixel, meaning that an instruction in an image is converted into a pixel.

In the present specification, a raster pattern updating method is provided, and the present specification relates to a raster pattern updating apparatus, a model training method, a model training apparatus, a computing device, and a computer-readable storage medium, which are described in detail in the following embodiments one by one.

Referring to the schematic diagram shown in fig. 1, in order to achieve the goal of concentrating limited resources into a region of interest and improving the target detection effect without increasing time consumption, the grid pattern updating method provided by the embodiment may first obtain a reference grid pattern including a target object and a region corresponding to the region of interest, and then determine a target region corresponding to the target object in the reference grid pattern, so as to construct a target distribution heat map corresponding to the reference grid pattern according to attribute information of the target region, thereby implementing the goal of reflecting the region of interest in the reference grid pattern through the heat map; meanwhile, position information of a target object is determined in the reference grid graph, so that an object route graph corresponding to the reference grid graph is constructed according to the position information, and a route area concerned in the reference grid graph is embodied through the route graph; at this time, considering that the area affecting the traveling of the target object is the area of interest and the route area, the target distribution heat map and the target route map can be used as the basis for dividing the reference grid map, so that when grid updating is performed, denser grids can be added for the area of interest and the route area, and sparser grids can be added for other areas; that is, a grid distribution curve is constructed based on the target distribution heat map and the target road map, and the attention in each direction is expressed by the curve, so that the reference grid map is updated to the target grid map. When the grid division is carried out, the focused content can be selected according to the running condition at the current moment to carry out dense grid division processing, and other areas adopt sparse grid division processing, so that the dynamic grid is realized, and the aims of balancing calculation resources and detecting accuracy are achieved.

Referring to fig. 2, fig. 2 shows a flowchart of a grid map updating method according to an embodiment of the present disclosure, which specifically includes the following steps.

Step S202, a reference raster image including a target object and an object of interest is acquired.

The grid map updating method can be applied to any unmanned scenes, such as scenes of unmanned private vehicles, unmanned public transport vehicles, unmanned delivery vehicles and the like; in this embodiment, the unmanned delivery vehicle is taken as an example to describe the grid map updating method, and descriptions in other scenes can be referred to the same or corresponding descriptions in this embodiment, which are not repeated here.

Specifically, the target object specifically refers to an object moving in a current unmanned scene, such as an unmanned delivery vehicle, a car, a transport vehicle and the like; accordingly, the object of interest specifically refers to an object of interest in the current unmanned scene, such as pedestrians, roadblocks, vehicles, green belts, and the like. And grid re-dividing processing is required to be carried out on a reference grid graph containing a target object and an object of interest in the current scene, so that the grid graph after re-dividing can be enabled to have the grid graph which is required to be used finally in a mode of adopting dense grids for an area of interest at the current moment, and other areas are displayed in a mode of adopting sparse grids, and therefore the calculation accuracy is ensured while the calculation resources can be reduced in the application stage. Correspondingly, the reference raster image specifically refers to a raster image which is generated for the target object at the current moment and is not subjected to raster updating, and the raster sizes of the reference raster image are equal, namely the reference raster image is a two-dimensional image with uniform grids, and the target raster image with non-uniform grid distribution can be obtained by carrying out dynamic raster division on the reference raster image, so that the target object is driven to run by utilizing the raster image of the mapping point cloud data in an unmanned scene.

Based on the above, in order to reduce the consumption of computing resources and improve the computing precision corresponding to the attention area, the reference raster image including the target object and the attention object may be acquired first, so as to locate the reference object and the attention object in the reference raster image, and then the grid update may be performed according to the positions and the attributes of the reference object and the attention object, so that the grid distribution of the attention area is denser, and the grid distribution of the non-attention area is sparser, thereby being more practical when the target object is driven to travel.

Further, when the reference raster pattern including the reference object and the object of interest is acquired, since the reference raster pattern is a raster pattern which is required to be subjected to raster repartition subsequently, and the base of the repartition is established on the object of interest and the reference object, the determination of the reference raster pattern needs to be completed by combining the geographic position information of the two objects; in this embodiment, the specific implementation manner is as follows:

loading a target virtual map according to the geographic position information of the target object, and constructing an initial grid map according to the target virtual map; updating the initial raster image based on the geographic position information of the target object and the geographic position information of the concerned object, and generating a reference raster image containing the target object and the concerned object according to an updating result; and the grids corresponding to the grids contained in the reference grid graph have the same size.

Specifically, the geographic position information specifically refers to position coordinates corresponding to the current moment of the object in the virtual map, and is used for positioning the position of the object in the map, so as to construct a reference raster pattern according to the target virtual map with the position of the object. The target virtual map specifically refers to a map used by a target object in the driving process, and the map has a mapping relationship with a driving area of the target object. Correspondingly, the initial raster image specifically refers to a two-dimensional image which is constructed by combining the target virtual map and is not subjected to raster division, and meanwhile, the initial raster image also does not contain target objects and attention objects.

Based on the above, in order to construct a target raster pattern in which the grids of the region of interest are dense and the grids of other regions are sparse, a target virtual map required to be used for constructing the raster pattern may be loaded according to the geographic position information of the target object, and then an initial raster pattern which does not include the target object and the target object of interest may be constructed according to the target virtual map. On this basis, in order to be able to construct a raster image that can be raster repartitioned using the initial raster image, it is necessary to add an object of interest and a target object to the initial raster image, thereby providing a basis for repartitioning the raster image. At this time, the initial grid graph can be updated according to the geographic position information of the target object and the geographic position information of the object of interest, so as to generate a reference grid graph containing the target object and the object of interest according to an updating result; and the grids contained in the reference grid graph are equal in size and are used for subsequent reconstruction.

For example, the unmanned delivery vehicle a for express delivery travels on the rightmost lane on the first road, during the traveling process, a virtual map used during the traveling process may be loaded according to the current positioning information of the unmanned delivery vehicle a, so as to construct an initial grid map according to the virtual map, and then, in combination with the current positioning information of the unmanned delivery vehicle a and the positioning information corresponding to the roadblock and the pedestrian required to be concerned during the traveling of the current unmanned delivery vehicle a, a reference grid map is constructed, the constructed reference grid map is as shown in fig. 3 (a), the grid size is uniform, and the unmanned delivery vehicle a, the roadblock and the pedestrian are included, so that the subsequent grid repartition is convenient for the roadblock and the pedestrian for use when the unmanned delivery vehicle a is driven to travel.

In summary, by constructing the reference raster graphs which have the same raster size and contain the target object and the target object, the follow-up construction of elements of the repartitioning raster graphs by using the target object and the reference object in the reference raster graphs can be facilitated, so that the actual running requirements can be combined to carry out raster repartitioning, and the calculation requirements in the actual scene can be met.

Step S204, determining a target area corresponding to the object of interest in the reference grid graph, and constructing a target distribution heat graph corresponding to the reference grid graph according to attribute information of the target area.

Specifically, after the reference grid map including the object of interest and the target object is obtained, further, considering that the grid repartition needs to depend on the content corresponding to the area of interest of the target object, such as a roadblock or a pedestrian occurring in front of the target object in running, if the target object needs to be driven to make an avoidance decision, the calculation needs to be completed by depending on the grid position of the point cloud mapping in the grid map, if the size of each grid is the same, the calculation resources will be uniformly distributed on each grid, so that the corresponding calculation resources may be insufficient for the grids corresponding to the point cloud mapping position, therefore, in order to improve the calculation resource utilization rate and the calculation accuracy of the area of interest in the grid map, the area of the associated object of interest may be divided by adopting dense grids, and other areas may be divided by adopting sparse grids. On the basis, in order to achieve the purpose of dividing grids, the concerned area corresponding to the concerned object can be determined in the reference grid graph, so that the target distribution heat graph corresponding to the reference grid graph is constructed according to the concerned area, the target distribution heat graph can be used as the basis of grid division subsequently, and therefore grid division corresponding to the concerned area is more dense in the grid graph.

The target area specifically refers to a target frame for selecting an object of interest in the reference raster image, and the target frame can be any polygonal frame, so long as the object of interest can be selected, the target frame can be used for reflecting the position of the object of interest in the raster image, and the aim of denser raster division of the area can be achieved when raster updating is performed. Accordingly, the attribute information specifically refers to attribute description information corresponding to the target area, including, but not limited to, the size, coordinates, and/or radius of the target area. Correspondingly, the target distribution heat map specifically refers to a heat map which is constructed based on attribute information of a target area and has a mapping relation with a reference grid map, the target distribution heat map comprises a plurality of heat areas, each heat area corresponds to a unique heat value, the height of the heat value is used for reflecting whether the heat area is related to an object of interest or not, and the heat area is used as grid distribution, so that grid division corresponding to the heat area with high heat value can be more densely performed when grid updating is performed.

Further, when the target distribution heat map is constructed, the target distribution heat map is considered to be used as grid distribution for updating the reference grid map, so that the updated grid map shows the effects of dense grids of the concerned region and sparse other regions; therefore, when the target distribution heat map is constructed, the region heat value can be calculated by combining the attribute information of the target region, and the method is used for realizing the embodiment of the position of the region of interest by utilizing the heat value, so that the method is used in the grid division stage; in this embodiment, the specific implementation manner is as follows:

Determining attribute information of the target area, and constructing an initial distribution heat map corresponding to the reference grid map; adding a region heat value for a heat region corresponding to the target region in the initial distribution heat map according to the attribute information, and adding a background heat value for a background region except the heat region; and generating the target distribution heat map according to the region heat value and the adding result of the background heat value.

Specifically, the initial distribution heat map is a heat map constructed based on a reference grid map, the heat map and the reference grid map have a mapping relationship, the initial distribution heat map comprises a plurality of areas, and a heat value corresponding to each area needs to be calculated according to attribute information of a target area and is used for reflecting the position of the target area, so that the heat map is mapped on the grid map for grid division. Correspondingly, the heat map region specifically refers to a region corresponding to the target region in the initial distribution heat map; correspondingly, the region heat value is specifically a heat value added to the heat region by a pointer, the heat value is added based on attribute information of a target region, the target region is used for representing the corresponding target region of the heat region, and the target region further comprises an object of interest, so that the heat region in the heat map can be identified by combining the heat value in the use stage, and the heat value is used for improving the density of the corresponding grid of the region during grid division.

The background region specifically refers to other heat regions in the initial distribution heat map, which do not correspond to the target region, because the partial region does not contain any object of interest, and still needs to be subjected to grid division, but only the grid division is relatively sparse, so that a background heat value can be added for the partial region. If a heat value is not added to the partial area, that is, if the heat value defaults to 0, a grid distribution curve constructed later is not smooth enough, so that grids in the repartitioned grid graph are too sparse or even have no grids, and the use of the grid graph is affected. The background heat value can be understood as a paranoid value of the heat value, and the density of the paranoid grid makes grid division more reasonable. In practical applications, the setting of the background heat value may be added according to the actual requirement, and the background heat value is not greater than the area heat value, which is not limited in this embodiment.

Based on this, in order to ensure that the grid division can be completed according to the region corresponding to the object of interest, and that the region grid division corresponding to the object of interest is dense, other region grids can be used while being sparse, therefore, the attribute information of the target region can be determined first, and an initial distribution heat map corresponding to the reference grid map can be constructed; at this time, an area heat value may be added to a heat area corresponding to the target area in the initial distribution heat map according to the attribute information, so as to reflect the position of the object of interest in the heat map. Meanwhile, in order to avoid that the heat value of other areas is in a default state and influences the accuracy of grid division, namely, the area is too sparse or too dense, the background heat value can be added for the background area outside the heat value area; the method and the device ensure that each region has a corresponding proper heat value, namely, a target distribution heat map is generated according to the adding result of the region heat value and the background heat value, and the target distribution heat map is convenient to use as grid distribution in the follow-up process.

In practical application, when the target distribution heat map is constructed, in order to ensure that the heat value of each region is endowed more reasonably and can be used as grid distribution, the size of a target frame corresponding to an object of interest in a reference grid map can be read, the Gaussian kernel radius is calculated according to the width and the height in the size, the radius is used as the embodiment of the heat value, a target distribution heat map (HeatMap) with the value distributed between 0 and 1 is manufactured, and the heat value in the target distribution heat map can be mapped with grid density, so that the heat value is used as grid distribution for subsequent grid updating, and the grid division of the region corresponding to the object of interest in the finally obtained target grid map is more dense. Meanwhile, in order to avoid the problem that the grid division of other areas is unreasonable, such as too sparse or no grid, the heat value of the background area needs to be improved, so that the problem that the background area is not allocated with grids is avoided. In addition, when grid distribution calculation is performed, mapping can be performed in combination with the distribution of the target frame corresponding to the object of interest, specifically the following formula (1):

D _v ＝max(D _o +η,1)； (1)

wherein D is _v Representing a grid distribution; d (D) _o Representing a distribution of the target boxes; η is a value between 0 and 1, i.e. a smoothing factor, calculated from the heat values in the heat map. The larger η, the smoother the resulting grid distribution, so that when a grid update is performed, the resulting grid is more Is dense for reflecting the position of the object of interest in the raster pattern.

In addition, since the target object is in a driving state, the calculated raster images of each frame corresponding to the target object may be different, and the position corresponding to the object of interest in each frame may also be changed, so that in order to determine the target area corresponding to the object of interest in the reference raster image, the target area of the object of interest in the current frame may be predicted by combining the raster images of the previous frame; in this embodiment, the specific implementation manner is as follows:

acquiring a history grid chart, and determining a history target area corresponding to the object of interest in the history grid chart; and adding the target region to the target object in the reference grid graph according to the region information of the historical target region.

Specifically, the historical grid graph specifically refers to a grid graph used in the previous frame of calculation, and correspondingly, the historical target area is a target area corresponding to the object of interest in the historical grid graph; the corresponding region information specifically refers to region information corresponding to the history target region, including but not limited to region size, position, and the like.

Based on this, in order to be able to accurately locate the target area of the object of interest, the target area of the current frame may be predicted in combination with the historical target area of the historical frame, i.e.: firstly, acquiring a history grid chart, and determining a history target area corresponding to an object of interest in the history grid chart; at this time, the position of the target region corresponding to the current frame can be predicted in the reference raster image aiming at the object of interest according to the region information of the historical target region, and the target distribution heat image can be conveniently constructed and used subsequently by adding the predicted position corresponding to the target region in the raster image corresponding to the current frame.

Along the above example, after obtaining the reference raster pattern as shown in fig. 3 (a), an initial distribution heat map corresponding to the reference raster pattern may be constructed at this time based on the target frames corresponding to the roadblocks and pedestrians in the reference raster pattern. And then, according to the sizes of the target frames of the roadblocks and the pedestrians, calculating the Gaussian kernel radius, and according to the calculation result, determining the heat value of the heat area corresponding to the roadblocks and the pedestrians in the initial distribution heat map. Meanwhile, a preset background heat value is added for background areas except roadblocks and pedestrians, so that a target distribution heat map corresponding to a reference grid map and with heat values distributed between 0 and 1 can be obtained, grid distribution calculation is carried out on the basis of the target distribution heat map, the areas with dense grid distribution can correspond to the roadblocks and pedestrians, and the target distribution heat map is convenient to use when the unmanned distribution vehicle A is driven to run.

In summary, by combining the position of the object of interest in the reference raster pattern to locate the target area and constructing the heat pattern based on the target area, the area corresponding to the target area can be embodied in the heat pattern, and subsequent raster distribution calculation based on the target area can be realized, so that the raster distribution corresponding to the object of interest can be promoted to be denser in the raster distribution, and other background areas are thinner, thereby enabling the updated target raster pattern to be more matched with road conditions in the real scene.

And step S206, determining the position information of the target object in the reference grid graph, and constructing an object route graph corresponding to the reference grid graph according to the position information.

Specifically, while the target distribution heat map is constructed, the region of interest in consideration of the road condition where the target object is currently traveling is more important than the target object, namely, the road distribution, that is, the motor vehicle lane, the non-motor vehicle lane, the zebra stripes, the yellow/white solid lines, and the like in the road. The part of road distribution also needs to correspond to dense grids so as to ensure the updated target grid diagram and drive the target objects to run safely and in compliance with traffic rules. Therefore, the object route map corresponding to the reference grid map can be constructed by combining the position information of the target object, and the road condition is embodied by the object route map, so that when grid re-dividing is performed, grids which are more densely distributed in the road area which is concerned in the road are obtained.

The location information specifically refers to location information corresponding to the target object in actual running, and is used for determining road information associated with the current running scene according to the location information after determining the location information, and for constructing an object route map according to the road information. Correspondingly, the object RoadMap is specifically a RoadMap (RoadMap) constructed after loading road information according to the position information, and is used for representing the road condition corresponding to the current position through the object RoadMap and representing the concerned road area, so that when grid division is performed, the grid corresponding to the concerned road area can be divided into a denser grid, and other areas are divided into a sparser grid.

Further, when the object roadmap is constructed, the object roadmap is considered to be used as grid distribution and is used for updating the reference raster map, so that the updated raster map shows the effects of dense grids of the concerned road area and sparse other areas; therefore, when constructing the object roadmap, the object roadmap may be constructed in combination with the route attention, and in this embodiment, the specific implementation manner is as follows:

acquiring position information corresponding to the target object, and loading object association information according to the position information; constructing an initial roadmap corresponding to the reference raster pattern, and determining an association region in the initial roadmap based on the object association information; adding route attention values for the associated areas and non-route attention values for non-associated areas other than the associated areas in the initial roadmap according to the object associated information; and generating the object roadmap according to the route attention value and the addition result of the non-route attention value.

Specifically, the object association information specifically refers to information corresponding to a road object that affects the running of the target object in a road area of the associated target object in the map, including, but not limited to, identification information, size information, position information, and the like of the road object, such as a crosswalk, a length m, and the like; correspondingly, the initial route map specifically refers to a route map which corresponds to the reference grid map and is not added with route attention values; correspondingly, the associated area specifically refers to an area associated with the road object in the initial roadmap; correspondingly, the non-associated region specifically refers to a region in the initial roadmap, which is not associated with the road object. Accordingly, the route attention value is specifically a set value added by a pointer to an associated area and is used for representing a road object to be focused in the driving process, the non-route attention value is specifically a set value added by a pointer to a non-associated area and is used for ensuring that the non-attention area does not influence the driving, but the attention value is also required to be set and is used for ensuring that the area can have grids with default sizes when the grid division is influenced, and the problem of no grids or sparse grids does not occur.

Based on the above, in order to ensure that the grid division can be completed according to the associated area corresponding to the road object concerned in the road, and the grid division of the associated area is dense, the grids of other areas can be sparse and can be used at the same time, the position information corresponding to the target object can be acquired according to the reference grid map, and at the moment, the object associated information can be loaded according to the position information; meanwhile, an initial route map corresponding to the reference grid map is constructed, then an association region is determined in the initial route map according to the object association information, on the basis, route attention values are added to the association region and non-route attention values are added to non-association regions except the association region according to the object association information, so that route attention values capable of promoting grid density are added to the association region, and non-route attention values capable of ensuring that grids exist but are not too dense are added to the non-association region; and then, according to the addition results of the route attention and the non-route attention, generating an object roadmap, and conveniently using the object roadmap as grid distribution.

In practical application, when the object roadmap is constructed, in order to ensure that the attention value of each area is more reasonable and can be used as grid distribution, the position of the current target object can be positioned, so that road information, namely information such as a travelable area, a crosswalk, a tide lane, a bus lane and the like in a road, is acquired from the map according to positioning information; and then setting the attention value of the concerned region to be 1 in the automatic driving scene, such as setting 1 in the travelable region and the crosswalk region, setting 0 in other regions, such as green belts and blind roads, setting 0, and adding the attention values of the associated region and the non-associated region in the initial roadmap based on the setting, so as to form an object roadmap, thereby facilitating the subsequent determination of grid distribution according to the attention value of each region in the object roadmap. In addition, when grid distribution calculation is performed, mapping may be performed in combination with the distribution of the associated area and the non-associated area, specifically the following formula (2):

D _v ＝max(D _r +η,1)； (2)

Wherein D is _v Representing a grid distribution; d (D) _r Representing a distribution of the associated regions; η is a value between 0 and 1, i.e. a smoothing factor, calculated from the values of interest in the roadmap; the larger eta, the smoother the distribution of the obtained grids, so that the obtained grids are denser when the grids are updated, and the grids are used for reflecting the positions of the road objects in the road in the grid diagram.

Along the above example, after obtaining the reference raster pattern as shown in fig. 3 (a), the position of the unmanned delivery vehicle a in the map may be first located at this time, so as to load the road information corresponding to the current position according to the position, and at the same time, construct the initial route pattern corresponding to the reference raster pattern. And then, determining a drivable area corresponding to the right lane in the initial roadmap according to the road information, setting a concerned value corresponding to the area as g1, setting concerned values of other areas as g2, and generating the roadmap according to the setting result, and calculating grid distribution based on the road map, wherein the areas with dense grid distribution correspond to the drivable area, so that the drivable area is convenient to use when the unmanned distribution vehicle A is driven to run.

In summary, by determining the concerned region and the non-concerned region in the initial roadmap in combination with the object association information and setting different concerned values for different regions, the distinction between the concerned region and the non-concerned region can be embodied in the finally constructed object roadmap, so that the density in grid distribution is affected, and the finally updated target raster graph can achieve the purposes of balancing calculation resources and calculation precision.

And step S208, constructing a grid distribution curve based on the target distribution heat map and the object roadmap, and updating the reference grid map into a target grid map according to the grid distribution curve.

Specifically, after the object roadmap and the target distribution heat map corresponding to the reference raster pattern are constructed, further, considering that the object roadmap and the target distribution heat map are constructed based on the attention area and the position information, and the content of interest, such as the content of pedestrians, crosswalks and the like, in the process of driving the target object can be reflected, so that in order to allocate more computing resources to the content of interest, a denser grid can be divided for the part of the content, and a sparser grid can be divided for other areas. Therefore, the grid distribution curve can be constructed based on the target distribution diagram and the object road diagram to realize that the distribution of the concerned region and the road region in the current driving scene is reflected through the grid distribution curve, and the distribution is used as the grid distribution to update the grids in the reference grid diagram, so that the reference grid diagram can be updated into the target grid diagram, and the method is convenient to use in the automatic driving scene.

The grid distribution curve specifically refers to a curve of corresponding grid distribution constructed according to a target distribution heat map and an object roadmap, the curve is used for reflecting probability of a heat map area, a background area and/or an associated area and an unassociated area in the target distribution heat map in the grid distribution direction, and the grid distribution density is reflected through the probability, so that when grid updating is carried out, the updating of grids can be completed according to the curve, the grid distribution of a concerned area is dense, and grids of other areas are sparse. Correspondingly, the target grid graph specifically refers to a grid graph obtained by updating a grid of the reference grid graph, and grid distribution in the grid graph is distributed in a distribution mode that the grid distribution of the target grid graph is dense according to the distribution of the region of interest and the grids of other regions are sparse.

Further, when the target distribution heat map and the object road map are utilized to construct a grid distribution curve and update the reference grid map, two different curves can be constructed by considering that the target distribution heat map corresponds to an object of interest and the object road map corresponds to road information, and the calculation of the grid size is carried out on the basis of the two different curves and is used for updating the grid map; in this embodiment, the specific implementation manner is as follows:

Constructing a target grid distribution curve according to the target distribution heat map, and constructing a route grid distribution curve according to the object route map; calculating grid size information based on the target grid profile and the route grid profile; and adjusting the grid size contained in the reference grid graph according to the grid size information, and generating a target grid graph according to an adjustment result.

Specifically, the target grid distribution curve is a grid distribution curve constructed based on a target distribution diagram, and is used for reflecting the probability of a heat map area and a background area in the grid distribution direction in the target distribution heat map, and reflecting the grid distribution density through the probability. Correspondingly, the route grid distribution curve specifically refers to a grid distribution curve constructed based on the object roadmap, and is used for reflecting the probability of the associated area and the non-associated area in the object roadmap in the grid distribution direction, and reflecting the grid distribution density through the probability. Correspondingly, the grid size information specifically refers to the grid size calculated by combining the target grid distribution curve and the route grid distribution curve, and in the calculation process, the size of each grid can be calculated by combining the two grid distribution curves in sequence, so that after all the grid sizes are calculated, the grid size can be adjusted by using the calculated grid size information.

When calculating the grid size information based on the target grid distribution curve and the route grid distribution curve, considering that the same grid may have probability values on both curves, namely the object of interest and the road object may overlap, calculating the size of one grid based on the two probability values at this time, and completing the calculation in a mean value taking mode or selecting the maximum value mode; alternatively, the two grid sizes may be calculated separately, and then the average value may be calculated, or the maximum value may be taken as the grid size, which is not limited in this embodiment.

Based on this, when updating the reference grid map, a target grid distribution curve can be constructed according to the target distribution heat map first, and a route grid distribution curve can be constructed according to the object route map; calculating grid size information by combining the target grid distribution curve and the route grid distribution curve; the grid size of each updated grid can be obtained, and then the grid size contained in the reference grid graph can be adjusted according to the grid size information, so that the target grid graph can be generated according to the adjustment result.

For example, after the heat map and the roadmap are obtained, a distribution curve of the target frame corresponding to the roadblock in the heat map in the x direction and the y direction can be constructed according to the heat value in the heat map, as shown in (b) of fig. 3, p1 in the curve represents the probability of each position of the roadblock in the x direction and the y direction, and in order to achieve the grid distribution, the grid distribution and the roadblock distribution are consistent according to the probability distribution of the roadblock, the calculation of the grid distribution can be performed based on the curve. At this time, the size of the grid corresponding to the roadblock can be calculated by using the curve, that is, the size of the grid 11 is s11×s11, the size of the grid 12 is s12×s12 … …, and the size of the grid 1n is s1n×s1n. Similarly, according to the attention value in the roadmap, a distribution curve of the travelable road region in the x-direction and the y-direction in the roadmap is constructed, as shown in fig. 3 (b), p2 in the curve represents the probability of each position of the travelable road region in the x-direction and the y-direction, and in order to achieve the arrangement of the grid distribution according to the probability distribution of the travelable road region, that is, the grid distribution is identical to the travelable road region distribution, the calculation of the grid distribution may be performed based on the curve. In this case, the size of the grid corresponding to the travelable road area may be calculated by using the curve, that is, the size of the grid 21 is s21×s21, the size of the grid 22 is s22×s22 … …, and the size of the grid 2n is s2n×s2n. After the sizes of the grids are obtained, the reference grid graph can be updated based on the grid sizes, and the target grid graph shown in fig. 4 is obtained according to the updating result, namely, the grid distribution corresponding to the roadblocks and the travelable roads is denser, and the grid distribution of other areas is more sparse.

In sum, by combining the target grid distribution curve and the route grid distribution curve to calculate the grid size information, the grid distribution can be influenced based on the region of interest, so that the grid distribution corresponding to the region of interest is more densely distributed, the grid distribution of other regions is more sparsely distributed, and the purpose of balancing calculation resources and accuracy is achieved.

Further, when the grid distribution curve is obtained to update the reference grid graph to the target grid graph, the grid scales in two directions are determined, and the grid is re-divided on the reference grid graph according to the scales, so that the target grid graph is obtained, and in the embodiment, the specific implementation manner is as follows:

constructing transverse distribution parameters of the related transverse dimension and longitudinal distribution parameters of the longitudinal dimension according to the grid distribution curve; generating a transverse grid scale based on the transverse distribution parameters and a longitudinal grid scale based on the longitudinal distribution parameters; and carrying out grid division on the reference grid graph according to the transverse grid scales and the longitudinal grid scales to obtain a target grid graph.

Specifically, the transverse distribution parameter specifically refers to a grid distribution parameter of an associated transverse dimension, and is used for reflecting the grid dimension proportion of the grid in the transverse dimension; correspondingly, the longitudinal distribution parameter specifically refers to a grid distribution parameter of an associated longitudinal dimension, and is used for representing the grid dimension of the grid in the longitudinal dimension. Correspondingly, the transverse grid scale and the longitudinal grid scale specifically refer to scales used for dividing grids, and are used for realizing grid re-division according to the scales, so that grid division efficiency can be ensured.

Based on the above, when grid division is performed according to the grid distribution curve, the transverse distribution parameters of the related transverse dimension and the longitudinal distribution parameters of the longitudinal dimension can be constructed according to the grid distribution curve; thereafter, a transverse grid scale may be generated based on the transverse distribution parameters, and a longitudinal grid scale may be generated based on the longitudinal distribution parameters; after the grid scales are obtained, grid division processing can be carried out on the reference grid graph according to the transverse grid scales and the longitudinal grid scales, so that the target grid graph is obtained according to division results.

In practical application, after the grid distribution curve is obtained, the grid distribution is mapped along the transverse direction and the longitudinal direction, namely, when the transverse scale is calculated, the longitudinal distribution parameters are compressed, and a one-dimensional vector of the related transverse dimension is obtained to calculate the transverse grid scale; similarly, when the longitudinal scale is calculated, the transverse distribution parameters are compressed, and a one-dimensional vector of the relevant longitudinal dimension is obtained to calculate the longitudinal grid scale.

Specifically, the discrete distribution can be fitted based on a grid distribution curve, then high-density interpolation is carried out to return to a discrete state, the area under the curve is calculated in a numerical calculation mode such as a trapz function, and finally the area between every two scales is approximate to one w of the longitudinal plane by searching the scales, wherein w is the total number of grids. Therefore, the grid scale meeting the use requirement can be obtained for grid division so as to obtain the target grid graph. In the specific implementation, when grid scale calculation is performed, the calculation process can be approximated by adopting a mode of predicting grid distribution by using a neural network, so that the calculation efficiency is improved.

In conclusion, the grid calculation efficiency can be effectively improved by adopting the grid distribution curve to calculate the distribution parameters in two dimensions and determining the grid scale on the basis of the distribution parameters, so that the updating accuracy of the target grid graph is ensured.

In addition, after the target raster image is obtained, the target raster image is considered to be the raster image for driving the target object to travel, so that after the point cloud data is acquired, the point cloud data needs to be mapped to the raster image for calculation and use of automatic driving. However, since the grid division becomes dense according to the position distribution of the object of interest and the road object, other areas are sparse, the grid coordinates are changed, so that the point cloud data cannot be mapped into the corresponding grids in the grid map, and in order to avoid the problem, the coordinate mapping relationship can be established first, so that the mapping can be completed according to the coordinate mapping relationship. In this embodiment, the specific implementation manner is as follows:

collecting point cloud data aiming at the target object, and constructing a coordinate mapping relation between the point cloud data and grids in the target grid graph; and mapping the point cloud data to the target raster pattern based on the coordinate mapping relation, wherein the target raster pattern mapped with the point cloud data is used for the target object to execute an object adjustment task.

Specifically, the point cloud data specifically refers to a three-dimensional point data set of the actual road surface acquired by a three-dimensional laser scanning instrument, wherein the three-dimensional points include, but are not limited to, three-dimensional points corresponding to any object on the road, such as vehicles, people, trees, roadblocks and the like; correspondingly, the coordinate mapping relationship is specifically a mapping relationship between point cloud coordinates in the point cloud data and grid coordinates of grids in the target grid graph, and each grid coordinate may correspond to one or more point cloud coordinates. Correspondingly, the object adjustment task specifically refers to a task of driving the target object to perform route adjustment, including but not limited to driving the target object at a constant speed, accelerating, changing lanes, braking, and the like.

Based on the above, as the grids contained in the target raster pattern change, the position coordinates of each grid change, if the uniform raster pattern is still used for the point cloud rasterization processing, the point cloud data cannot be mapped to the real position, so that the driving safety of the target object is affected. Therefore, in order to ensure that the point cloud data can still be accurately mapped into the corresponding grid, after the point cloud data is acquired, the grid coordinates are repositioned for the non-uniform grid in the grid graph, and then a coordinate mapping relation is established between the point cloud data and the grid with the repositioned grid coordinates, so that the point cloud data can be accurately mapped into the target grid graph for subsequent processing according to the coordinate mapping relation. And then, the point cloud data can be mapped into the target raster pattern based on the coordinate mapping relation, so that the target raster pattern mapped with the point cloud data is obtained, and the safe driving of the target object is realized. And the target raster image of the mapping point cloud data drives the target object to execute the object adjustment task.

For example, after the three-dimensional laser scanning instrument configured for the unmanned distribution vehicle a collects the point cloud data, the point cloud coordinate and the grid coordinate can be matched, so that the point cloud coordinate x1 and the grid coordinate b1 can be determined to be matched according to the matching result; the point cloud coordinate x2 is matched with the grid coordinate b1, the … … point cloud coordinate x is matched with the grid coordinate b, and after all matching relations are determined, the point cloud coordinate can be used for mapping the point cloud coordinate into a target grid graph later, so that the unmanned distribution vehicle A can safely and normally run. For example, when a roadblock occurs in the front, the unmanned delivery vehicle A can be driven to change the road leftwards, and the like.

In sum, the mapping of the point cloud data into the target raster pattern is completed by establishing the coordinate mapping relation, so that the mapping result is identical to the position relation of the real scene, and the safe and stable running of the target object can be ensured.

The following describes, with reference to fig. 5, an example of application of the raster image update method provided in the present specification in an unmanned scenario. Fig. 5 shows a flowchart of a processing procedure of a grid map updating method according to an embodiment of the present disclosure, which specifically includes the following steps.

Step S502, loading a target virtual map according to geographic position information of a target vehicle, and constructing an initial grid map according to the target virtual map;

step S504, updating the initial raster pattern based on the geographic position information of the target vehicle and the geographic position information of the vehicle of interest, and generating a reference raster pattern including the target vehicle and the vehicle of interest based on the update result.

In step S506, a target area corresponding to the vehicle of interest is determined in the reference raster pattern, attribute information of the target area is determined, and an initial distribution heat map corresponding to the reference raster pattern is constructed.

In step S508, a region heat value is added to the heat region corresponding to the target region in the initial distribution heat map according to the attribute information, and a background heat value is added to the background region outside the heat region.

Step S510, generating a target distribution heat map according to the adding result of the regional heat value and the background heat value.

Step S512, position information corresponding to the target vehicle is obtained, and vehicle association information is loaded according to the position information.

Step S514, an initial route pattern corresponding to the reference grid pattern is constructed, and the association area is determined in the initial route pattern based on the vehicle association information.

Step S516, adding route attention to the associated area and non-route attention to the non-associated area other than the associated area in the initial route map according to the vehicle-associated information.

Step S518, a vehicle route map is generated according to the addition results of the route attention and the non-route attention.

Step S520, constructing a target grid distribution curve according to the target distribution heat map, and constructing a route grid distribution curve according to the vehicle route map.

Step S522, calculating grid size information based on the target grid distribution curve and the route grid distribution curve.

Step S524, adjusting the grid size included in the reference grid pattern according to the grid size information, and generating a target grid pattern according to the adjustment result.

Thereafter, point cloud data can be collected for the target vehicle, and a coordinate mapping relationship between the point cloud data and grids in the target grid graph can be constructed; and mapping the point cloud data to a target raster pattern based on the coordinate mapping relation, wherein the target raster pattern mapped with the point cloud data is used for a target vehicle to execute a vehicle adjustment task.

In order to achieve the goal of concentrating limited resources into a region of interest and improving the target detection effect without increasing time consumption, the grid map updating method provided by the embodiment can firstly acquire a reference grid map containing a target vehicle and a corresponding target region of interest, and then determine the target region corresponding to the target vehicle in the reference grid map, so as to construct a target distribution heat map corresponding to the reference grid map according to attribute information of the target region, and embody the region of interest in the reference grid map through the heat map; meanwhile, position information of a target vehicle is determined in the reference grid graph, so that a vehicle route graph corresponding to the reference grid graph is constructed according to the position information, and a route area concerned in the reference grid graph is embodied through the route graph; at this time, considering that the area affecting the running of the target vehicle is the area of interest and the route area, the target distribution heat map and the vehicle route map can be used as the basis for dividing the reference grid map, so that when grid updating is performed, denser grids can be added for the area of interest and the route area, and sparser grids can be added for other areas; that is, a grid distribution curve is constructed based on the target distribution heat map and the vehicle road map, and the attention in each direction is expressed by the curve, so that the reference grid map is updated to the target grid map. When the grid division is carried out, the focused content can be selected according to the running condition at the current moment to carry out dense grid division processing, and other areas adopt sparse grid division processing, so that the dynamic grid is realized, and the aims of balancing calculation resources and detecting accuracy are achieved.

Corresponding to the above method embodiments, the present disclosure further provides another embodiment of a method for updating a raster image, and fig. 6 shows a schematic diagram of another method for updating a raster image provided in one embodiment of the present disclosure. As shown in fig. 6, the method is applied to a vehicle control end, and includes:

step S602, acquiring a reference grid pattern including a target vehicle and a vehicle of interest;

step S604, determining a target area corresponding to the concerned vehicle in the reference grid graph, and constructing a target distribution heat graph corresponding to the reference grid graph according to attribute information of the target area;

step S606, determining the position information of the target vehicle in the reference grid graph, and constructing an object route graph corresponding to the reference grid graph according to the position information;

and step S608, constructing a grid distribution curve based on the target distribution heat map and the object road map, and updating the reference grid map into a target grid map according to the grid distribution curve, wherein the target grid map is used for automatic driving of the target vehicle.

The other grid map updating method provided in this embodiment corresponds to the grid map updating method in the above embodiment, and the same or similar description contents can be referred to the corresponding contents in the above embodiment, which is not repeated here. The vehicle control end is a terminal device for controlling the target vehicle to run.

In summary, in order to achieve the goal of concentrating the limited resources into the region of interest, and improving the target detection effect without increasing time consumption, the target region corresponding to the target vehicle and the reference grid graph corresponding to the interest may be first obtained, and then the target region corresponding to the target vehicle is determined in the reference grid graph, so as to construct a target distribution heat map corresponding to the reference grid graph according to the attribute information of the target region, and implement the goal of reflecting the region of interest in the reference grid graph through the heat map; meanwhile, position information of a target vehicle is determined in the reference grid graph, so that a vehicle route graph corresponding to the reference grid graph is constructed according to the position information, and a route area concerned in the reference grid graph is embodied through the route graph; at this time, considering that the area affecting the running of the target vehicle is the area of interest and the route area, the target distribution heat map and the vehicle route map can be used as the basis for dividing the reference grid map, so that when grid updating is performed, denser grids can be added for the area of interest and the route area, and sparser grids can be added for other areas; that is, a grid distribution curve is constructed based on the target distribution heat map and the vehicle road map, and the attention in each direction is expressed by the curve, so that the reference grid map is updated to the target grid map. When the grid division is carried out, the focused content can be selected according to the running condition at the current moment to carry out dense grid division processing, and other areas adopt sparse grid division processing, so that the dynamic grid is realized, and the aims of balancing calculation resources and detecting accuracy are achieved.

Corresponding to the method embodiments described above, the present disclosure further provides a model training method embodiment, and fig. 7 shows a schematic diagram of a model training method provided in one embodiment of the present disclosure. As shown in fig. 7, the method includes:

step S702, obtaining a reference raster pattern comprising a target object and an object of interest, and inputting the reference raster pattern into an initial raster division model for processing to obtain predicted raster size information;

step S704, determining a target area corresponding to the object of interest in the reference grid graph, and constructing a target distribution heat graph corresponding to the reference grid graph according to attribute information of the target area;

step S706, determining the position information of the target object in the reference raster pattern, and constructing an object route pattern corresponding to the reference raster pattern according to the position information;

step S708, constructing a grid distribution curve based on the target distribution heat map and the object roadmap, and determining reference grid size information according to the grid distribution curve;

step S710, calculating a loss value based on the predicted grid size information and the reference grid size information, and tuning the initial grid division model using the loss value until a target grid division model satisfying a training stop condition is obtained.

Specifically, the initial raster division model is a model capable of performing raster division processing on the content presented in the reference raster pattern according to the raster size, wherein the input of the model is the reference raster pattern, the output is raster size information, and the raster size information corresponds to the characteristics of dense regions of interest and other region coefficients. Correspondingly, the predicted grid size information specifically refers to a predicted result obtained after the sample is processed by the initial grid division model in a training stage. Correspondingly, the reference grid size information specifically refers to a label corresponding to the sample, that is, the real grid size information, and the calculation of the size information is the same as that in the above embodiment or the corresponding description, which is not repeated herein. Accordingly, the training stopping condition specifically refers to a condition for stopping training the initial grid division model, including but not limited to an iteration number condition, a loss value comparison condition, a verification set verification condition, and the like.

Based on the above, in order to quickly complete the grid division in the application stage, the grid division model may be trained in advance, and during training, a sample, that is, a reference grid pattern including the object of interest and the target object may be acquired first, and input into the initial grid division model to be trained for processing, so as to obtain the predicted grid size information. Meanwhile, the reference grid size information is calculated as a label by using the grid division method in the above embodiment. And at the moment, calculating a loss value by using the label and the model prediction result, judging whether the model after parameter adjustment meets the training stop condition after parameter adjustment is carried out on the model by using the loss value, if not, continuously selecting a new sample, and repeating the process until a target grid division model meeting the training stop condition is obtained. In the application stage, the target grid graph with the divided grids can be obtained directly by using the trained model and used for driving the object to run.

Corresponding to the method embodiment, the present disclosure further provides a raster image update apparatus embodiment, and fig. 8 shows a schematic structural diagram of the raster image update apparatus according to one embodiment of the present disclosure. As shown in fig. 8, the apparatus includes:

an acquisition module 802 configured to acquire a reference raster pattern including a target object and an object of interest;

a first construction module 804, configured to determine a target area corresponding to the object of interest in the reference grid graph, and construct a target distribution heat map corresponding to the reference grid graph according to attribute information of the target area;

a second construction module 806 configured to determine position information of the target object in the reference raster pattern, and construct an object roadmap corresponding to the reference raster pattern according to the position information;

an updating module 808 is configured to construct a grid distribution curve based on the target distribution heat map and the object roadmap, and update the reference grid map to a target grid map according to the grid distribution curve.

In an alternative embodiment, the first building block 804 is further configured to:

In an alternative embodiment, the second building block 806 is further configured to:

In an alternative embodiment, the update module 808 is further configured to:

In an alternative embodiment, the apparatus further comprises:

the adding module is configured to acquire a history grid chart and determine a history target area corresponding to the attention object in the history grid chart; and adding the target region to the target object in the reference grid graph according to the region information of the historical target region.

In an alternative embodiment, the obtaining module 802 is further configured to:

In an alternative embodiment, the apparatus further comprises:

the rasterization module is configured to acquire point cloud data aiming at the target object and construct a coordinate mapping relation between the point cloud data and grids in the target raster image; and mapping the point cloud data to the target raster pattern based on the coordinate mapping relation, wherein the target raster pattern mapped with the point cloud data is used for the target object to execute an object adjustment task.

In order to achieve the goal of concentrating limited resources into a region of interest and improving the target detection effect without increasing time consumption, the grid map updating device provided by the embodiment may first obtain a reference grid map including a target object and a region of interest corresponding to the target object, and then determine a target region corresponding to the target object in the reference grid map, so as to construct a target distribution heat map corresponding to the reference grid map according to attribute information of the target region, and implement the goal of reflecting the region of interest in the reference grid map through the heat map; meanwhile, position information of a target object is determined in the reference grid graph, so that an object route graph corresponding to the reference grid graph is constructed according to the position information, and a route area concerned in the reference grid graph is embodied through the route graph; at this time, considering that the area affecting the traveling of the target object is the area of interest and the route area, the target distribution heat map and the target route map can be used as the basis for dividing the reference grid map, so that when grid updating is performed, denser grids can be added for the area of interest and the route area, and sparser grids can be added for other areas; that is, a grid distribution curve is constructed based on the target distribution heat map and the target road map, and the attention in each direction is expressed by the curve, so that the reference grid map is updated to the target grid map. When the grid division is carried out, the focused content can be selected according to the running condition at the current moment to carry out dense grid division processing, and other areas adopt sparse grid division processing, so that the dynamic grid is realized, and the aims of balancing calculation resources and detecting accuracy are achieved.

The above is a schematic scheme of a raster image update apparatus of the present embodiment. It should be noted that, the technical solution of the grid map updating apparatus and the technical solution of the grid map updating method belong to the same concept, and details of the technical solution of the grid map updating apparatus, which are not described in detail, can be referred to the description of the technical solution of the grid map updating method.

Corresponding to the above method embodiment, the present disclosure further provides another embodiment of a raster image update apparatus, and fig. 9 shows a schematic structural diagram of another raster image update apparatus provided in one embodiment of the present disclosure. As shown in fig. 9, the device is applied to a vehicle control terminal, and includes:

an acquire raster pattern module 902 configured to acquire a reference raster pattern including a target vehicle and a vehicle of interest;

a heat map constructing module 904 configured to determine a target area corresponding to the vehicle of interest in the reference grid map, and construct a target distribution heat map corresponding to the reference grid map according to attribute information of the target area;

a build roadmap module 906 configured to determine positional information of the target vehicle in the reference raster pattern and build an object roadmap corresponding to the reference raster pattern according to the positional information;

An update grid map module 908 is configured to construct a grid distribution curve based on the target distribution heat map and the object roadmap, and update the reference grid map as a target grid map according to the grid distribution curve, wherein the target grid map is used for automatic driving of the target vehicle.

The above is another exemplary embodiment of the raster image update apparatus of the present embodiment. It should be noted that, the technical solution of the grid map updating apparatus and the technical solution of the grid map updating method belong to the same concept, and details of the technical solution of the grid map updating apparatus, which are not described in detail, can be referred to the description of the technical solution of the grid map updating method.

Corresponding to the method embodiment, the present disclosure further provides an embodiment of a model training device, and fig. 10 shows a schematic structural diagram of the model training device provided in one embodiment of the present disclosure. As shown in fig. 10, the apparatus includes:

the model processing module 1002 is configured to acquire a reference raster pattern including a target object and an object of interest, and input the reference raster pattern to an initial raster division model for processing, so as to obtain predicted raster size information;

A distribution heat map constructing module 1004, configured to determine a target area corresponding to the object of interest in the reference grid map, and construct a target distribution heat map corresponding to the reference grid map according to attribute information of the target area;

a build object roadmap module 1006 configured to determine location information of the target object in the reference raster pattern and build an object roadmap corresponding to the reference raster pattern according to the location information;

a determine size information module 1008 configured to construct a grid distribution curve based on the target distribution heat map and the object roadmap, and determine reference grid size information from the grid distribution curve;

the model tuning module 1010 is configured to calculate a loss value based on the predicted grid size information and the reference grid size information, and tune the initial grid division model using the loss value until a target grid division model satisfying a training stop condition is obtained.

The above is a schematic scheme of a model training apparatus of the present embodiment. It should be noted that, the technical solution of the model training device and the technical solution of the model training method belong to the same concept, and details of the technical solution of the model training device, which are not described in detail, can be referred to the description of the technical solution of the model training method.

It should be noted that, according to the types of the automatic driving vehicles, the algorithm modules (the functional modules) mentioned in the above embodiments may also be different. For example, different algorithm modules may be involved for logistics vehicles, public service vehicles, medical service vehicles, terminal service vehicles. The algorithm module is illustrated below for each of these four autonomous vehicles:

the logistics vehicles refer to vehicles used in logistics scenes, and can be logistics vehicles with automatic sorting functions, logistics vehicles with refrigerating and heat-preserving functions and logistics vehicles with measuring functions. These logistics vehicles may involve different algorithm modules.

For example, for a logistics vehicle, an automated sorting device may be provided which can automatically pick up and transport, sort and store goods after the logistics vehicle arrives at the destination. This involves an algorithm module for sorting of goods, which mainly implements logic control of goods taking out, handling, sorting and storing.

For another example, for a cold chain logistics scene, the logistics vehicle can be further provided with a refrigeration and heat preservation device, and the refrigeration and heat preservation device can realize refrigeration or heat preservation of transported fruits, vegetables, aquatic products, frozen foods and other perishable foods, so that the fruits, vegetables, aquatic products, frozen foods and other perishable foods are in a proper temperature environment, and the problem of long-distance transportation of perishable foods is solved. The algorithm module is mainly used for dynamically and adaptively calculating proper temperature of cold food or heat preservation according to information such as food (or article) properties, perishability, transportation time, current seasons, weather and the like, and automatically adjusting the cold food or heat preservation device according to the proper temperature, so that transportation personnel do not need to manually adjust the temperature when different foods or articles are transported by a vehicle, the transportation personnel are liberated from complicated temperature regulation and control, and the efficiency of cold food or heat preservation transportation is improved.

For example, in most logistics scenes, the charge is carried out according to the volume and/or weight of the packages, the number of the logistics packages is very large, and the volume and/or weight of the packages are simply measured by an express delivery person, so that the efficiency is very low, and the labor cost is high. Therefore, in some logistics vehicles, a measuring device is additionally arranged, so that the volume and/or the weight of the logistics package can be automatically measured, and the cost of the logistics package can be calculated. This involves an algorithm module for logistic parcel measurement which is primarily used to identify the type of logistic parcel, determine the way in which the logistic parcel is measured, such as whether a volumetric measurement or a weight measurement is made or a combination of volumetric and weight measurements are made simultaneously, and can perform volumetric and/or weight measurements based on the determined way of measurement, and perform cost calculations based on the measurement results.

The public service vehicle refers to a vehicle for providing a certain public service, and can be, for example, a fire truck, a deicing vehicle, a sprinkler, a snow plow, a garbage disposal vehicle, a traffic guidance vehicle and the like. These public service vehicles may involve different algorithm modules.

For example, for an automatically driven fire engine, the main task is to perform a reasonable fire extinguishing task for a fire scene, which involves an algorithm module for the fire extinguishing task, and the algorithm module at least needs to implement logic of fire condition identification, fire extinguishing scheme planning, automatic control of a fire extinguishing device and the like.

For another example, for deicing vehicles, the main task is to remove ice and snow on the road surface, which involves an algorithm module for deicing that at least needs to implement logic for identifying ice and snow conditions on the road surface, making deicing schemes based on the ice and snow conditions, such as which road segments need to be defrosted, which road segments need not be defrosted, whether salt spraying mode, salt spraying gram number, etc. are used, and automatic control of the deicing device in case of determining the deicing scheme.

The medical service vehicle is an automatic driving vehicle capable of providing one or more medical services, and the vehicle can provide medical services such as disinfection, temperature measurement, medicine preparation, isolation and the like, and the medical service vehicle relates to algorithm modules for providing various self-service medical services, wherein the algorithm modules mainly realize the identification of disinfection requirements and the control of disinfection devices so as to enable the disinfection devices to disinfect patients or identify the positions of the patients, control the temperature measurement devices to automatically measure the temperature of the patients at the positions of the forehead and the like of the patients, or realize the judgment of symptoms, give medicine according to the judgment result and need to realize the identification of medicines/medicine containers, control the medicine taking mechanical arm so as to enable the medicine taking mechanical arm to take medicines for the patients according to the medicine prescription, and the like.

The terminal service vehicle refers to a self-service type automatic driving vehicle capable of replacing some terminal equipment to provide certain convenience services for users, for example, the vehicle can provide printing, attendance checking, scanning, unlocking, payment, retail and other services for the users.

For example, in some application scenarios, users often need to go to a particular location to print or scan a document, which is time consuming and laborious. Therefore, there is a terminal service vehicle capable of providing a printing/scanning service for a user, the service vehicles can be interconnected with a user terminal device, the user sends a printing command through the terminal device, the service vehicle responds to the printing command, automatically prints a document required by the user and can automatically send the printed document to a user position, the user does not need to go to a printer for queuing, and the printing efficiency can be greatly improved. Or, the user can respond to the scanning instruction sent by the terminal equipment and move to the user position, and the user can finish scanning on the scanning tool of the service vehicle for placing the document to be scanned, so that queuing at a printer/scanner is not needed, and time and labor are saved. This involves an algorithm module providing print/scan services that at least needs to identify interconnections with the user terminal device, responses to print/scan instructions, positioning of user location, travel control, etc.

For another example, as new retail projects develop, more and more electronic commerce uses self-service vending machines to sell goods to various office buildings and public areas, but the self-service vending machines are placed at fixed positions and are not movable, and users need to go to the self-service vending machines before they can purchase the required goods, so that convenience is still poor. The self-service driving vehicles capable of providing retail services are arranged, the service vehicles can bear goods to automatically move, corresponding self-service shopping APP or shopping portals can be provided, a user can place an order to the self-service driving vehicles providing retail services through the APP or shopping portals by means of terminals such as mobile phones, the order comprises names, quantity and user positions of goods to be purchased, after receiving an order placing request, the vehicles can determine whether the current remaining goods have the goods purchased by the user and whether the quantity is enough, and under the condition that the goods purchased by the user are determined to be enough, the goods can be carried to the user positions automatically, and the goods are provided for the user, so that the convenience of shopping of the user is further improved, the user time is saved, and the user can use the time for more important things. This involves the algorithm modules providing retail services that implement mainly logic for responding to user order requests, order processing, merchandise information maintenance, user location positioning, payment management, etc.

Fig. 11 illustrates a block diagram of a computing device 1100 provided according to one embodiment of the present description. The components of computing device 1100 include, but are not limited to, a memory 1110 and a processor 1120. Processor 1120 is coupled to memory 1110 via bus 1130, and database 1150 is used to hold data.

The computing device 1100 also includes an access device 1140, the access device 1140 enabling the computing device 1100 to communicate via one or more networks 1160. Examples of such networks include public switched telephone networks (PSTN, public Switched Telephone Network), local area networks (LAN, local Area Network), wide area networks (WAN, wide Area Network), personal area networks (PAN, personal Area Network), or combinations of communication networks such as the internet. The access device 1140 may include one or more of any type of network interface, wired or wireless, such as a network interface card (NIC, network interface controller), such as an IEEE802.11 wireless local area network (WLAN, wireless Local Area Network) wireless interface, a worldwide interoperability for microwave access (Wi-MAX, worldwide Interoperability for Microwave Access) interface, an ethernet interface, a universal serial bus (USB, universal Serial Bus) interface, a cellular network interface, a bluetooth interface, near field communication (NFC, near Field Communication).

In one embodiment of the present description, the above components of computing device 1100, as well as other components not shown in FIG. 11, may also be connected to each other, such as by a bus. It should be understood that the block diagram of the computing device illustrated in FIG. 11 is for exemplary purposes only and is not intended to limit the scope of the present description. Those skilled in the art may add or replace other components as desired.

Computing device 1100 may be any type of stationary or mobile computing device, including a mobile computer or mobile computing device (e.g., tablet, personal digital assistant, laptop, notebook, netbook, etc.), mobile phone (e.g., smart phone), wearable computing device (e.g., smart watch, smart glasses, etc.), or other type of mobile device, or a stationary computing device such as a desktop computer or personal computer (PC, personal Computer). Computing device 1100 may also be a mobile or stationary server.

Wherein the processor 1120 is configured to execute computer-executable instructions that, when executed by the processor, perform the steps of the grid map updating method or the model training method described above.

The foregoing is a schematic illustration of a computing device of this embodiment. It should be noted that, the technical solution of the computing device and the technical solution of the above grid map updating method or the model training method belong to the same concept, and details of the technical solution of the computing device, which are not described in detail, can be referred to the description of the technical solution of the above grid map updating method or the model training method.

An embodiment of the present disclosure also provides a computer-readable storage medium storing computer-executable instructions that, when executed by a processor, implement the steps of the above-described grid map updating method or model training method.

The above is an exemplary version of a computer-readable storage medium of the present embodiment. It should be noted that, the technical solution of the storage medium and the technical solution of the above grid map updating method or the model training method belong to the same concept, and details of the technical solution of the storage medium which are not described in detail can be referred to the description of the technical solution of the above grid map updating method or the model training method.

An embodiment of the present disclosure further provides a computer program, where the computer program, when executed in a computer, causes the computer to perform the steps of the above-mentioned grid map updating method or model training method.

The above is an exemplary version of a computer program of the present embodiment. It should be noted that, the technical solution of the computer program and the technical solution of the above grid map updating method or model training method belong to the same concept, and details of the technical solution of the computer program, which are not described in detail, can be referred to the description of the technical solution of the above grid map updating method or model training method.

The foregoing describes specific embodiments of the present disclosure. Other embodiments are within the scope of the following claims. In some cases, the actions or steps recited in the claims can be performed in a different order than in the embodiments and still achieve desirable results. In addition, the processes depicted in the accompanying figures do not necessarily require the particular order shown, or sequential order, to achieve desirable results. In some embodiments, multitasking and parallel processing are also possible or may be advantageous.

The computer instructions include computer program code that may be in source code form, object code form, executable file or some intermediate form, etc. The computer readable medium may include: any entity or device capable of carrying the computer program code, a recording medium, a U disk, a removable hard disk, a magnetic disk, an optical disk, a computer Memory, a Read-Only Memory (ROM), a random access Memory (RAM, random Access Memory), an electrical carrier signal, a telecommunications signal, a software distribution medium, and so forth. It should be noted that the content of the computer readable medium can be increased or decreased appropriately according to the requirements of the patent practice, for example, in some areas, according to the patent practice, the computer readable medium does not include an electric carrier signal and a telecommunication signal.

It should be noted that, for simplicity of description, the foregoing method embodiments are all expressed as a series of combinations of actions, but it should be understood by those skilled in the art that the embodiments are not limited by the order of actions described, as some steps may be performed in other order or simultaneously according to the embodiments of the present disclosure. Further, those skilled in the art will appreciate that the embodiments described in the specification are all preferred embodiments, and that the acts and modules referred to are not necessarily all required for the embodiments described in the specification.

In the foregoing embodiments, the descriptions of the embodiments are emphasized, and for parts of one embodiment that are not described in detail, reference may be made to the related descriptions of other embodiments.

The preferred embodiments of the present specification disclosed above are merely used to help clarify the present specification. Alternative embodiments are not intended to be exhaustive or to limit the invention to the precise form disclosed. Obviously, many modifications and variations are possible in light of the teaching of the embodiments. The embodiments were chosen and described in order to best explain the principles of the embodiments and the practical application, to thereby enable others skilled in the art to best understand and utilize the invention. This specification is to be limited only by the claims and the full scope and equivalents thereof.

Claims

1. A raster image update method, comprising:

2. The grid map updating method according to claim 1, wherein the constructing a target distribution heat map corresponding to the reference grid map according to the attribute information of the target region includes:

determining attribute information of the target area, and constructing an initial distribution heat map corresponding to the reference grid map;

adding a region heat value for a heat region corresponding to the target region in the initial distribution heat map according to the attribute information, and adding a background heat value for a background region except the heat region;

And generating the target distribution heat map according to the region heat value and the adding result of the background heat value.

3. The grid map updating method according to claim 1, wherein the determining the position information of the target object in the reference grid map and constructing an object route map corresponding to the reference grid map according to the position information includes:

acquiring position information corresponding to the target object, and loading object association information according to the position information;

constructing an initial roadmap corresponding to the reference raster pattern, and determining an association region in the initial roadmap based on the object association information;

adding route attention values for the associated areas and non-route attention values for non-associated areas other than the associated areas in the initial roadmap according to the object associated information;

and generating the object roadmap according to the route attention value and the addition result of the non-route attention value.

4. The grid map updating method according to claim 1, wherein the constructing a grid distribution curve based on the target distribution heat map and the object roadmap, and updating the reference grid map to a target grid map in accordance with the grid distribution curve, comprises:

Constructing a target grid distribution curve according to the target distribution heat map, and constructing a route grid distribution curve according to the object route map;

calculating grid size information based on the target grid profile and the route grid profile;

and adjusting the grid size contained in the reference grid graph according to the grid size information, and generating a target grid graph according to an adjustment result.

5. The grid map updating method according to claim 1, wherein the updating the reference grid map to the target grid map according to the grid distribution curve comprises:

constructing transverse distribution parameters of the related transverse dimension and longitudinal distribution parameters of the longitudinal dimension according to the grid distribution curve;

generating a transverse grid scale based on the transverse distribution parameters and a longitudinal grid scale based on the longitudinal distribution parameters;

and carrying out grid division on the reference grid graph according to the transverse grid scales and the longitudinal grid scales to obtain a target grid graph.

6. The grid map updating method according to claim 1, wherein before the step of determining the target area corresponding to the object of interest in the reference grid map is performed, the method further comprises:

Acquiring a history grid chart, and determining a history target area corresponding to the object of interest in the history grid chart;

and adding the target region to the target object in the reference grid graph according to the region information of the historical target region.

7. The grid map updating method according to any one of claims 1 to 6, the acquiring a reference grid map including a target object and an object of interest, comprising:

loading a target virtual map according to the geographic position information of the target object, and constructing an initial grid map according to the target virtual map;

updating the initial raster image based on the geographic position information of the target object and the geographic position information of the concerned object, and generating a reference raster image containing the target object and the concerned object according to an updating result;

and the grids corresponding to the grids contained in the reference grid graph have the same size.

8. The grid map updating method according to any one of claims 1 to 6, further comprising, after the step of updating the reference grid map to the target grid map according to the grid distribution curve is performed:

collecting point cloud data aiming at the target object, and constructing a coordinate mapping relation between the point cloud data and grids in the target grid graph;

And mapping the point cloud data to the target raster pattern based on the coordinate mapping relation, wherein the target raster pattern mapped with the point cloud data is used for the target object to execute an object adjustment task.

9. A grid map updating method is applied to a vehicle control end and comprises the following steps:

10. A model training method, comprising:

11. A computing device, comprising:

a memory and a processor;

the memory is configured to store computer executable instructions, the processor being configured to execute the computer executable instructions, which when executed by the processor, implement the steps of the method of any one of claims 1 to 10.

12. A computer readable storage medium storing computer executable instructions which when executed by a processor implement the steps of the method of any one of claims 1 to 10.