WO2022063056A1

WO2022063056A1 - Method for constructing high-precision point cloud map

Info

Publication number: WO2022063056A1
Application number: PCT/CN2021/119050
Authority: WO
Inventors: 彭国旗; 黄友; 张国龙; 张放; 李晓飞; 张德兆; 王肖; 霍舒豪
Original assignee: 重庆兰德适普信息科技有限公司
Priority date: 2020-09-22
Filing date: 2021-09-17
Publication date: 2022-03-31
Also published as: CN112414415A; CN112414415B

Abstract

A method for constructing a high-precision point cloud map, comprising: receiving first sensor data sent by a terminal (101) ; determining whether the first sensor data meets a data eligibility determination rule (104); when the first sensor data meets the data eligibility determination rule, obtaining second sensor data according to sensor data (105); constructing a point cloud map according to the second sensor data (106), the point cloud map comprising first positioning data; and when the point cloud map meets accuracy conditions, distortion conditions, smoothness conditions, continuity conditions and complete conditions, determining that the point cloud map passes verification (112). The method for constructing the high-precision point cloud map improves the quality of the point cloud map, and may construct multiple point cloud map tasks at the same time. Also provided are a device, a computer-readable storage medium, a computer server and a mobile tool.

Description

High-precision point cloud map construction method

This application claims the priority of the Chinese patent application with application number 202011003627.4 and titled "Method for constructing high-precision point cloud maps", filed on September 22, 2020, the disclosure of which is incorporated by reference This article.

technical field

The invention relates to the technical field of automatic driving, in particular to a method for constructing a high-precision point cloud map.

Background technique

At present, the most widely used point cloud map construction methods in the autonomous driving industry can be roughly divided into two categories: point cloud map construction methods based on offline data and point cloud map construction methods based on online data.

The point cloud map construction method based on offline data, if the point cloud map construction is completed offline based on the vehicle, not only the performance of the vehicle processor is very high, the large map is difficult to process, the map quality and error cannot be guaranteed, and it is impossible to complete multiple tasks at the same time. Point cloud map construction; if the point cloud map construction is completed offline based on the local desktop computer, it involves tedious processes such as data download, software startup, and result data upload, which requires more human intervention, and the degree of automation is low, resulting in reduced efficiency, and it is impossible to achieve multiple The scene point cloud map is constructed at the same time, which seriously wastes labor resources and local computing resources.

The point cloud map construction method based on online data, since the point cloud map construction tasks are all completed in the on-board processor, the performance of the on-board processor is very high, the large map is difficult to process, the map quality cannot be guaranteed, and the point cloud map verification is missing. Later data accumulation is not friendly, and multiple point cloud map construction tasks cannot be realized at the same time due to vehicle and scene limitations.

SUMMARY OF THE INVENTION

The purpose of the present invention is to provide a high-precision point cloud map construction method in view of the defects existing in the prior art. The method is executed by a cloud server, and can simultaneously realize the construction of multiple point cloud map tasks, which overcomes the difficulty in handling large maps. Difficulties. By verifying the constructed point cloud map, the quality of the point cloud map is improved.

In order to achieve the above purpose, in the first aspect, the present invention provides a method for constructing a high-precision point cloud map, and the method for constructing a high-precision point cloud map includes:

receiving first sensor data sent by the terminal; the first sensor data includes point cloud data, inertial navigation data, wheel speedometer data, GPS positioning data and image data;

After receiving the first sensor data, generating a map task construction request message for constructing a map;

According to the map task construction request message, obtain the current memory occupancy and the current map task construction number;

When the current memory occupancy is less than the preset memory occupancy, and the current map task construction number is less than the preset map task construction threshold, determine whether the first sensor data complies with the data eligibility judgment rule;

When the first sensor data complies with the data eligibility judgment rule, obtain second sensor data according to the point cloud data, inertial navigation data, wheel speedometer data, GPS positioning data and image data;

constructing a point cloud map according to the second sensor data; the point cloud map includes a plurality of points, each of which has first positioning data;

The point cloud map is verified. When the flag bit of the dynamic carrier phase difference technology RTK indicates that the signal is normal, the first positioning data is compared with the second positioning data obtained by the real-time dynamic carrier phase difference technology RTK. When the difference between the first positioning data and the second positioning data is not greater than a preset threshold, it is determined that the point cloud map meets the accuracy condition;

According to the inertial navigation data and the wheel speedometer data, a local path with a preset length is generated, and the local path is compared with the local path in the point cloud map of the same preset length. When the difference between the curvatures of the points of the local paths in the point cloud map of length is not greater than a preset difference threshold, it is determined that the point cloud map meets the distortion conditions;

Judging whether there is a jump in the connection between the multiple first positioning data in the point cloud map, when there is no jump, it is determined that the point cloud map meets the smoothness condition;

Determine whether the distance between any adjacent points corresponding to the first positioning data in the point cloud map is not greater than a preset distance threshold, and when it is not greater than a preset distance When the distance threshold is , it is determined that the point cloud map meets the continuity condition;

Determine whether the point is complete, and when the point is complete, determine that the point cloud map meets the complete condition;

When the point cloud map meets the accuracy conditions, distortion conditions, smoothness conditions, continuity conditions and integrity conditions, it is determined that the point cloud map passes the verification.

Preferably, the first sensor data sent by the receiving terminal specifically includes:

receiving multiple frames of raw sensor data sent by the terminal; each frame of the raw sensor data includes a frame header;

According to the frame header, determine whether the transmission of the multi-frame raw sensor data is completed;

When the transmission is completed, multiple frames of original sensor data are spliced according to the frame header to obtain the first sensor data.

Preferably, before the first sensor data sent by the receiving terminal includes:

The sensor receives a map collection instruction sent by the terminal, and collects data according to the map collection instruction according to a preset route and a preset manner;

When the data collection is completed, the terminal instructs the data collection to be completed through the export instruction.

Preferably, the judging whether the first sensor data complies with the data eligibility judging rule specifically includes:

judging whether the format of the first sensor data is a preset format;

When the format of the first sensor data does not conform to the preset format, a first prompt message for instructing to re-acquire the first sensor data is generated, and the first prompt message is sent to the terminal.

judging whether the data quality of the first sensor meets the requirements;

When the quality of the first sensor data does not meet the requirements, a second prompt message for instructing to re-acquire the original sensor data is generated, and the second prompt message is sent to the terminal.

Preferably, the point cloud data is a first data set corresponding to a first sampling frequency, and each data in the first data set has a first timestamp. When the first timestamp is inconsistent with the first sampling frequency , it is determined that the quality of the point cloud data does not meet the requirements; and/or,

The inertial navigation data is a second data set corresponding to the second sampling frequency, each data in the second data set has a second time stamp, and when the second time stamp is inconsistent with the second sampling frequency, determine the quality of the inertial navigation data is not satisfactory; and/or,

The wheel speedometer data is a third data set corresponding to a third sampling frequency, and each data in the third data set has a third time stamp. When the third time stamp is inconsistent with the third sampling frequency, determine that the quality of the wheel speedometer data is not satisfactory; and/or,

The GPS positioning data is a fourth data set corresponding to the fourth sampling frequency, and each data in the fourth data set has a fourth time stamp. When the fourth time stamp is inconsistent with the fourth sampling frequency, determine The quality of the GPS positioning data does not meet the requirements; and/or,

The image data is a fifth data set corresponding to the fifth sampling frequency, each data in the fifth data set has a fifth time stamp, and when the fifth time stamp is inconsistent with the fifth sampling frequency, it is determined that the The quality of the image data described above does not meet the requirements.

Preferably, the inertial navigation data includes acceleration information, angular velocity information and attitude angle information of the vehicle; the wheel speed meter data includes angular velocity information, linear velocity information and vehicle yaw rate information of the left and right wheels of the vehicle; point cloud data, inertial navigation data, wheel speed meter data, GPS positioning data and image data to obtain second sensor data, which specifically includes: combining the point cloud data, the acceleration information of the vehicle, the angular velocity information and the attitude angle information, The angular velocity information, linear velocity information and vehicle yaw rate information of the left and right wheels of the vehicle, the GPS positioning data and the first image data are fused to obtain second sensor data.

In a second aspect, the present invention provides a device including a memory and a processor, where the memory is used for storing a program, and the processor is used for executing the method for constructing a high-precision point cloud map according to any one of the first aspect.

In a third aspect, the present invention provides a computer-readable storage medium, where a computer program is stored on the computer-readable storage medium, and when the computer program is executed by a processor, the high precision described in any one of the first aspect is realized. Point cloud map construction method.

In a fourth aspect, the present invention provides a computer server, including: a memory, a processor, and a transceiver; the processor is configured to be coupled to the memory, read and execute instructions in the memory, to implement the first aspect The method for constructing a high-precision point cloud map; the transceiver is coupled to the processor, and the processor controls the transceiver to send and receive messages.

In a fifth aspect, the present invention provides a mobile tool, including the computer server described in the fourth aspect.

In the method for constructing a high-precision point cloud map provided by the embodiment of the present invention, the cloud server automatically responds to and triggers the point cloud map construction task, which can simultaneously realize the construction of multiple point cloud map tasks, and overcomes the difficulty that large maps are difficult to handle; When the first sensor data complies with the data eligibility judgment rules, the construction of the point cloud map will be carried out, and through the verification of the integrity, accuracy, distortion, smoothness and continuity of the point cloud map, the time and cost of constructing the point cloud map will be reduced. The error satisfies the high-precision requirement of point cloud map construction, thereby improving the quality of point cloud map.

Description of drawings

FIG. 1 is a schematic flowchart of a method for constructing a high-precision point cloud map according to Embodiment 1 of the present invention.

detailed description

The technical solutions of the present invention will be further described in detail below through the accompanying drawings and embodiments.

The method for constructing a high-precision point cloud map provided by the first embodiment of the present invention is applied to a cloud server for constructing a high-precision point cloud map.

It should be noted that the sensor data required for constructing a high-precision point cloud map is collected by an intelligent vehicle, where an intelligent vehicle can be understood as an unmanned autonomous vehicle. Smart vehicles are installed with a variety of sensors, and the source of point cloud map data is based on sensors, including but not limited to cameras, lidars, global positioning systems (GPS), inertial measurement units (Inertial measurement unit) , IMU) and wheel speedometer. Among them, there can be multiple cameras and lidars, which are respectively set around the smart vehicle; GPS is set on the top of the vehicle; the inertial measurement unit can be set in a hidden position such as the chassis of the vehicle; there can be two wheel speedometers, which are respectively set on the top of the vehicle. on the left and right wheels of the vehicle.

FIG. 1 is a schematic flowchart of a method for constructing a high-precision point cloud map according to Embodiment 1 of the present invention. As shown in FIG. 1 , the method for constructing a high-precision point cloud map provided by Embodiment 1 of the present invention includes the following steps:

Step 101: Receive first sensor data sent by a terminal; the first sensor data includes point cloud data, inertial navigation data, wheel speed meter data, GPS positioning data, and image data;

Specifically, the point cloud data is the data obtained by the laser radar, and the data of the laser radar can be one or more. When the number of the laser radar is multiple, the point cloud data represents the point cloud obtained by the multiple laser radars. collection of data.

Inertial navigation data refers to the current acceleration information, angular velocity information, and attitude angle information of the vehicle obtained through the inertial measurement unit IMU.

The wheel speedometer data is the angular velocity information, linear speed information and vehicle yaw rate information of the left and right wheels of the vehicle obtained by the wheel speedometer.

GPS positioning data is the position information of the vehicle acquired by GPS. Due to errors in GPS positioning, the location information of the vehicle is not the current precise location information of the vehicle. In a preferred embodiment, in order to reduce the error between the position information of the vehicle and the current position of the vehicle, the GPS is preferably differential GPS, and the intelligent vehicle obtains the position information of the vehicle detected by the differential GPS.

The image data is the environment data around the vehicle collected by the cameras. When the number of cameras is multiple, the image data represents a set of environment data collected by the multiple cameras.

Before step 101, the intelligent vehicle receives the driving information sent by the cloud server or the terminal, the driving information includes destination and time information, and the intelligent vehicle automatically drives to the corresponding block according to the destination information and time information. After receiving the map collection instruction sent by the terminal, the sensor installed on the smart vehicle collects data according to the map collection instruction, according to the preset route and preset method; when the sensor collects the data, the terminal will indicate the completion of the data collection through the export instruction. The map collection instruction may be input by the user through the terminal, or may be sent by the server to the terminal.

The first sensor data that the cloud server receives from the terminal specifically includes:

Receive multiple frames of raw sensor data sent by the terminal; each frame of raw sensor data includes a frame header;

According to the frame header, determine whether the multi-frame raw sensor data has been transmitted;

Due to the different sensor models, the input sensor data types and formats are different, which is not conducive to the universality of the algorithm. Therefore, it is necessary to preprocess the original sensor data, convert the sensor data into a specific data type and format, and implement the algorithm. Consistency of input data. Therefore, the first sensor data is uploaded to the cloud server after the original sensor data is sorted and perfected.

Step 102, after receiving the first sensor data, generate a map task construction request message for constructing a map;

Specifically, the cloud server will detect in real time whether there is uploading of the first sensor data, and at the same time monitor whether the uploading of the first sensor data is completed. When the upload of the first sensor data is completed, a map construction command will be triggered, and the cloud server will automatically assign a task name with a unique identification to the task corresponding to the map construction command, and automatically generate different map task construction request messages according to different task names. , so as to construct the map, so that the cloud server can automatically respond and trigger the point cloud map construction task.

Step 103, according to the map task construction request message, obtain the current memory occupancy and the current map task construction number;

Step 104, when the current memory occupancy is less than the preset memory occupancy, and the current map task construction number is less than the preset map task construction threshold, determine whether the first sensor data complies with the data eligibility judgment rule;

Specifically, the cloud server will set a number of configuration parameters, including the memory usage and the number of tasks.

The preset memory usage can be understood as the memory usage allowed by the server. In a specific example, the preset memory footprint can be set to 80%. The preset map task construction threshold can be understood as the number of tasks that can run simultaneously on the cloud server. In another specific example, the preset map task construction threshold is N, 0<N<6, and N is a positive integer .

The cloud server can judge whether the point cloud map construction task needs to wait for the completion of other tasks by judging the memory usage and the number of tasks running at the same time. In a specific example, when the memory usage is 60% and the number of tasks running simultaneously on the cloud server is 4, the cloud server can execute the point cloud map construction task without waiting. Therefore, the point cloud map construction through the cloud server can avoid the disadvantage of high requirements on the vehicle processor when the map is constructed through the vehicle processor in the existing online map construction technology, and can realize the simultaneous realization of multiple point cloud map tasks. Build, overcoming the intractable difficulties of a large map.

Step 105, when the first sensor data complies with the data eligibility judgment rule, obtain second sensor data according to point cloud data, inertial navigation data, wheel speedometer data, GPS positioning data and image data;

Specifically, the data eligibility judgment rule specifically includes whether the format of the first sensor data is a preset format and whether the quality of the first sensor data meets the requirements.

When the format of the first sensor data does not conform to the preset format, the cloud server will generate a first prompt message, and send the first prompt message to the terminal to instruct the terminal to re-acquire the first sensor data, that is, the original sensor data according to the preset format The format is rearranged and uploaded to the cloud server.

Further, since each sensor collects data according to a preset sampling frequency, it is determined whether the quality of the first sensor data meets the requirements by checking whether the data measured by a certain sensor is missing in the first sensor data.

The point cloud data is the first data set corresponding to the first sampling frequency, and each data in the first data set has a first timestamp. When the first timestamp is inconsistent with the first sampling frequency, the quality of the point cloud data is determined. does not meet the requirements; and/or, the inertial navigation data is a second data set corresponding to the second sampling frequency, and each data in the second data set has a second time stamp, when the second time stamp is inconsistent with the second sampling frequency , it is determined that the quality of the inertial navigation data does not meet the requirements; and/or, the wheel speedometer data is a third data set corresponding to the third sampling frequency, and each data in the third data set has a third time stamp, when the third time stamp When it is inconsistent with the third sampling frequency, it is determined that the data quality of the wheel speedometer does not meet the requirements; and/or, the GPS positioning data is a fourth data set corresponding to the fourth sampling frequency, and each data in the fourth data set has a fourth time Stamp, when the fourth time stamp is inconsistent with the fourth sampling frequency, it is determined that the quality of the GPS positioning data does not meet the requirements; and/or, the image data is the fifth data set corresponding to the fifth sampling frequency, and each data in the fifth data set Both have a fifth time stamp, and when the fifth time stamp is inconsistent with the fifth sampling frequency, it is determined that the quality of the image data does not meet the requirements.

When the quality of the first sensor data does not meet the requirements, the cloud server generates a second prompt message, and sends the second prompt message to the terminal to instruct the terminal to re-collect and acquire the original sensor data.

Therefore, only when the format of the first sensor data meets the preset format and the quality of the first sensor data meets the requirements, the cloud server will perform the point cloud map construction task. The setting of the data eligibility judgment rules greatly improves the quality of the sensor data for constructing the point cloud map, reduces the data error of the point cloud map construction, and improves the quality of the constructed point cloud map.

Step 106, constructing a point cloud map according to the second sensor data; the point cloud map includes a plurality of points, and each point has the first positioning data;

Specifically, the point cloud data, the acceleration information of the vehicle, the angular velocity information and the attitude angle information, the angular velocity information of the left and right wheels of the vehicle, the linear velocity information and the yaw rate information of the vehicle, the GPS positioning data and the first image data are fused to process, The second sensor data is obtained, that is, the second sensor data is the data after fusion processing of the first sensor data. The first positioning data is the positioning pose information of each point in the point cloud map calculated by a preset algorithm in the point cloud map, where the positioning pose information may include positioning information and an attitude angle.

Step 107, verify the point cloud map, when the flag bit of the dynamic carrier phase difference technology RTK indicates that the signal is normal, compare the first positioning data with the second positioning data obtained by the real-time dynamic carrier phase difference technology RTK, when the first positioning data When the difference between the first positioning data and the second positioning data is not greater than the preset threshold, it is determined that the point cloud map meets the accuracy condition;

Specifically, the cloud server performs quality verification on the above-mentioned successfully constructed point cloud map. In order to reduce the error of map data and improve the quality of map data, the constructed point cloud map must meet certain accuracy conditions. Therefore, in this method, in addition to using two differential GPS measurements, the real-time kinematic (RTK) carrier phase difference technology is also used to obtain the second positioning data. When the RTK flag bit of the dynamic carrier phase difference technology indicates that the signal is normal , compare the first positioning data with the second positioning data, determine whether the difference between the first positioning data and the second positioning data meets the preset threshold, and if the difference is not greater than the preset threshold, the point cloud map can be determined The accuracy meets the accuracy conditions, and provides centimeter-level positioning results to meet the requirements of high-precision point cloud map construction.

Step 108: Generate a local path with a preset length according to the inertial navigation data and the wheel speedometer data, and compare the local path with the local path in the point cloud map of the same preset length. When the difference between the curvatures of each point of the local path in the cloud map is not greater than the preset difference threshold, it is determined that the point cloud map meets the distortion condition;

Specifically, according to the timestamp, the inertial navigation data and the wheel speedometer data are aligned in time, and a certain preset length is selected to generate a local path. In a specific example, the preset length may be 10m. If the difference between the two aligned local paths (curves) is too large, it means that the shapes of the two curves are different. If they are in the same coordinate system, it means that one of them is distorted.

Step 109, judging whether there is a jump in the connection between the multiple first positioning data in the point cloud map, and when there is no jump, it is determined that the point cloud map meets the smoothness condition;

Specifically, smoothness means that the connection line of the points corresponding to the first positioning data in the constructed point cloud map should be a smooth curve. If a certain point jumps, there will be a jagged connection. Therefore, by judging whether there is a jump in the connection between the plurality of first positioning data, it can be judged whether the point cloud map meets the smoothness condition.

Step 110: Determine whether the distance between any adjacent points corresponding to two first positioning data in the point cloud map is not greater than a preset distance threshold, and when it is not greater than a preset distance threshold When , it is determined that the point cloud map meets the continuity condition;

Specifically, the distance between any two adjacent points corresponding to the first positioning data in the point cloud map has a preset distance threshold. Therefore, the adjacent points in the point cloud map are The distance between the points is compared with the preset distance threshold. When the distance is not greater than the preset distance threshold, it is considered that the continuity of the points is good, and the point cloud map meets the requirements of continuity. In a specific example, the preset distance threshold is 1m, and if the distance between two points is 2m, it can be determined that the continuity of the point cloud map does not meet the conditions.

Step 111, judge whether the point is complete, and when the point is complete, determine that the point cloud map meets the complete condition;

Specifically, the complete condition refers to judging whether the points in the point cloud map are complete, and the points include but are not limited to function points, traffic light identification data, and boundary information. Among them, the function point refers to the preset point on the point cloud map. For example, in the point cloud map on the user terminal, the selected point. The traffic light identification data refers to the position of the traffic light, such as the coordinates of the traffic light on the point cloud map, and the coordinates may be the position of the center point of the light board frame.

In one example, when judging that the boundary information is complete, the number of trajectory points on the boundary can be calculated according to the sensor frequency and the acquisition time of the sensor, and the boundary trajectory data in the generated point cloud map and the calculated boundary trajectory data on the boundary can be checked. Whether the number of track points is consistent. When the boundary trajectory data in the generated point cloud map is consistent with the calculated number of trajectory points on the boundary, it is determined that the boundary information satisfies the complete condition. When the function points, traffic light identification data and boundary information all meet the complete conditions, it is determined that the point cloud map meets the complete conditions.

Step 112, when the point cloud map meets the accuracy conditions, distortion conditions, smoothness conditions, continuity conditions and integrity conditions, it is determined that the point cloud map passes the verification.

In a preferred embodiment, when the point cloud map verification fails, for example, when the point cloud map does not meet the accuracy conditions, distortion conditions, smoothness conditions and continuity conditions, the cloud server will automatically modify the point cloud map construction needs. Specify parameters, such as positioning deviation, positioning status, etc., and re-execute the point cloud map construction with the new parameters, and then re-verify the point cloud map quality.

If the point cloud map does not meet the complete conditions, the cloud server will give specific data missing items, prompting the terminal to re-collect and upload the original data. Then re-execute steps 101-112 until the point cloud map quality meets the complete conditions.

Through the verification of the point cloud map, the error of the constructed point cloud map is reduced, thereby improving the quality of the point cloud map.

Further, the present application may further include: after the verification of the point cloud map is passed, storing the sensor data corresponding to the point cloud map, so that the sensor data corresponding to the stored multiple areas can be spliced to obtain a larger block of sensor data. Point cloud map.

The method for constructing a high-precision point cloud map provided by the first embodiment of the present invention, through the automatic response and triggering of the point cloud map construction by the cloud server, can realize the construction of multiple point cloud map tasks at the same time, and overcome the difficulty that the large map is difficult to handle; When the first sensor data complies with the data eligibility judgment rules, the construction of the point cloud map will be carried out, and through the verification of the integrity, accuracy, distortion, smoothness and continuity of the point cloud map, the time required for the constructed point cloud map will be reduced. The error satisfies the high-precision requirements of point cloud map construction, thereby improving the quality of the point cloud map.

The second embodiment of the present invention provides a device including a memory and a processor, where the memory is used to store a program, and the processor is used to execute the high-precision point cloud map construction method provided by the first embodiment of the present invention.

Embodiment 3 of the present invention provides a computer-readable storage medium, where a computer program is stored on the computer-readable storage medium, and when the computer program is executed by a processor, the method for constructing a high-precision point cloud map provided by Embodiment 1 of the present invention is implemented.

Embodiment 4 of the present invention provides a computer server, including: a memory, a processor, and a transceiver; the processor is configured to be coupled with the memory, and read and execute instructions in the memory, so as to implement the first embodiment of the present invention. The above-mentioned method for constructing a high-precision point cloud map; the transceiver is coupled to the processor, and the processor controls the transceiver to send and receive messages.

The fifth embodiment of the present invention provides a mobile tool, including the computer server described in the fourth embodiment. The mobile tool can be any movable tool, such as a vehicle (such as a vacuum cleaner, a sweeper, a floor scrubber, a logistics trolley, a passenger car, a sanitation vehicle, a bus, a coach, a van, a truck, a truck, Trailers, dump trailers, cranes, excavators, scrapers, road trains, sweepers, sprinklers, garbage trucks, engineering vehicles, rescue vehicles, AGVs (Automated Guided Vehicles), motorcycles, bicycles, Tricycles, hand carts, robots, sweepers, balance vehicles, etc., the application does not strictly limit the types of mobile tools, and will not be exhaustive here.

Professionals should be further aware that the units and algorithm steps of each example described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, computer software, or a combination of the two. Interchangeability, the above description has generally described the components and steps of each example in terms of function. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Skilled artisans may implement the described functionality using different methods for each particular application, but such implementations should not be considered beyond the scope of the present invention.

The steps of a method or algorithm described in connection with the embodiments disclosed herein may be implemented in hardware, a software module executed by a processor, or a combination of the two. Software modules can be placed in random access memory (RA high-precision point cloud map construction method), memory, read-only memory (RO high-precision point cloud map construction method), electrically programmable RO high-precision point cloud map construction method, electrically erasable A programmable RO high-precision point cloud map construction method, a register, a hard disk, a removable disk, a CD-RO high-precision point cloud map construction method, a power system control method, or any other form of storage medium known in the technical field.

The specific embodiments described above further describe the objectives, technical solutions and beneficial effects of the present invention in detail. It should be understood that the above descriptions are only specific embodiments of the present invention and are not intended to limit the scope of the present invention. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention shall be included within the protection scope of the present invention.

Claims

A method for constructing a high-precision point cloud map, characterized in that the method for constructing a high-precision point cloud map comprises:

receiving first sensor data sent by the terminal; the first sensor data includes point cloud data, inertial navigation data, wheel speedometer data, GPS positioning data and image data;

After receiving the first sensor data, generating a map task construction request message for constructing a map;

According to the map task construction request message, obtain the current memory occupancy and the current map task construction number;

When the current memory occupancy is less than the preset memory occupancy, and the current map task construction number is less than the preset map task construction threshold, determine whether the first sensor data complies with the data eligibility judgment rule;

When the first sensor data complies with the data eligibility judgment rule, obtain second sensor data according to the point cloud data, inertial navigation data, wheel speedometer data, GPS positioning data and image data;

constructing a point cloud map according to the second sensor data; the point cloud map includes a plurality of points, each of which has first positioning data;

The point cloud map is verified. When the flag bit of the dynamic carrier phase difference technology RTK indicates that the signal is normal, the first positioning data is compared with the second positioning data obtained by the real-time dynamic carrier phase difference technology RTK. When the difference between the first positioning data and the second positioning data is not greater than a preset threshold, it is determined that the point cloud map meets the accuracy condition;

According to the inertial navigation data and the wheel speedometer data, a local path with a preset length is generated, and the local path is compared with the local path in the point cloud map of the same preset length. When the difference between the curvatures of the points of the local paths in the point cloud map of preset length is not greater than a preset difference threshold, it is determined that the point cloud map meets the distortion condition;

Judging whether there is a jump in the connection between the plurality of the first positioning data in the point cloud map, and when there is no jump, it is determined that the point cloud map meets the smoothness condition;

Determine whether the distance between any adjacent points corresponding to the first positioning data in the point cloud map is not greater than a preset distance threshold, and when it is not greater than a preset distance When the distance threshold is , it is determined that the point cloud map meets the continuity condition;

Determine whether the point is complete, and when the point is complete, determine that the point cloud map meets the complete condition;

When the point cloud map meets the accuracy conditions, distortion conditions, smoothness conditions, continuity conditions and integrity conditions, it is determined that the point cloud map passes the verification.
The method for constructing a high-precision point cloud map according to claim 1, wherein the receiving the first sensor data sent by the terminal specifically includes:

receiving multiple frames of raw sensor data sent by the terminal; each frame of the raw sensor data includes a frame header;

According to the frame header, determine whether the transmission of the multi-frame raw sensor data is completed;

When the transmission is completed, multiple frames of original sensor data are spliced according to the frame header to obtain the first sensor data.
The method for constructing a high-precision point cloud map according to claim 1 or 2, wherein before the receiving the first sensor data sent by the terminal, the method further comprises:

The sensor receives a map collection instruction sent by the terminal, and collects data according to the map collection instruction according to a preset route and a preset manner;

When the data collection is completed, the terminal instructs the data collection to be completed through the export instruction.
The method for constructing a high-precision point cloud map according to any one of claims 1-3, wherein the judging whether the first sensor data complies with the data eligibility judgment rule specifically includes:

judging whether the format of the first sensor data is a preset format;

When the format of the first sensor data does not conform to the preset format, a first prompt message for instructing to re-acquire the first sensor data is generated, and the first prompt message is sent to the terminal.
The method for constructing a high-precision point cloud map according to any one of claims 1-4, wherein the judging whether the first sensor data complies with a data qualified judgment rule specifically includes:

judging whether the data quality of the first sensor meets the requirements;

When the quality of the first sensor data does not meet the requirements, a second prompt message for instructing to re-acquire the original sensor data is generated, and the second prompt message is sent to the terminal.
The method according to any one of claims 1-5, wherein the point cloud data is a first data set corresponding to a first sampling frequency, and each data in the first data set has a first timestamp, when the first timestamp is inconsistent with the first sampling frequency, it is determined that the quality of the point cloud data does not meet the requirements; and/or,

The inertial navigation data is a second data set corresponding to the second sampling frequency, each data in the second data set has a second time stamp, and when the second time stamp is inconsistent with the second sampling frequency, determine the quality of the inertial navigation data is not satisfactory; and/or,

The wheel speedometer data is a third data set corresponding to a third sampling frequency, and each data in the third data set has a third time stamp. When the third time stamp is inconsistent with the third sampling frequency, determine that the quality of the wheel speedometer data is not satisfactory; and/or,

The GPS positioning data is a fourth data set corresponding to the fourth sampling frequency, and each data in the fourth data set has a fourth time stamp. When the fourth time stamp is inconsistent with the fourth sampling frequency, determine The quality of the GPS positioning data does not meet the requirements; and/or,

The image data is a fifth data set corresponding to the fifth sampling frequency, each data in the fifth data set has a fifth time stamp, and when the fifth time stamp is inconsistent with the fifth sampling frequency, it is determined that the The quality of the image data described above does not meet the requirements.
The method for constructing a high-precision point cloud map according to any one of claims 1-6, wherein the inertial navigation data includes acceleration information, angular velocity information and attitude angle information of the vehicle; the wheel speed meter data includes Angular velocity information, linear velocity information and vehicle yaw rate information of the left and right wheels of the vehicle;

The obtaining of the second sensor data according to the point cloud data, inertial navigation data, wheel speedometer data, GPS positioning data and image data, specifically includes;

The point cloud data, the acceleration information of the vehicle, the angular velocity information and the attitude angle information, the angular velocity information of the left and right wheels of the vehicle, the linear velocity information and the vehicle yaw rate information, the GPS positioning data and all The first image data is fused to obtain second sensor data.
A device, characterized in that the device comprises a memory and a processor, the memory is used to store a program, and the processor is used to execute the method for constructing a high-precision point cloud map according to any one of claims 1-7 .
A computer-readable storage medium, characterized in that, a computer program is stored on the computer-readable storage medium, and when the computer program is executed by a processor, the high-precision point according to any one of claims 1-7 is realized Cloud map construction method.
A computer server, comprising: a memory, a processor and a transceiver;

The processor is configured to be coupled with the memory, and read and execute the instructions in the memory, so as to realize the method for constructing a high-precision point cloud map according to any one of claims 1-7;

The transceiver is coupled to the processor, and the processor controls the transceiver to send and receive messages.
A mobile tool, characterized in that it comprises the computer server of claim 10 .