CN109993700B

CN109993700B - Data processing method, device, electronic equipment and computer readable storage medium

Info

Publication number: CN109993700B
Application number: CN201910265528.4A
Authority: CN
Inventors: 罗盼; 白宇; 张振理; 王方伟
Original assignee: Beijing Baidu Netcom Science and Technology Co Ltd
Current assignee: Beijing Baidu Netcom Science and Technology Co Ltd
Priority date: 2019-04-03
Filing date: 2019-04-03
Publication date: 2023-07-04
Anticipated expiration: 2039-04-03
Also published as: CN109993700A

Abstract

The invention provides a data processing method, a data processing device, electronic equipment and a computer readable storage medium. The method comprises the following steps: according to the original point cloud, performing splicing treatment and fusion treatment to obtain a first treatment point cloud; receiving input operation of a user under the condition that the first processing point cloud does not meet a first preset condition; the input operation is used for designating homonymous feature points in the first processing point cloud; according to the homonymous feature points, carrying out fusion processing on the first processing point cloud to obtain a second processing point cloud; and under the condition that the second processing point cloud meets the first preset condition and the distance between the homonymous feature points in the second processing point cloud is smaller than the first preset distance, executing map production operation according to the second processing point cloud. Therefore, in the embodiment of the invention, the three-dimensional high-precision map can be produced according to the point cloud with the quality meeting the requirement, and compared with the prior art, the embodiment of the invention can effectively ensure the quality of the generated map data.

Description

Data processing method, device, electronic equipment and computer readable storage medium

Technical Field

The embodiment of the invention relates to the technical field of communication, in particular to a data processing method, a data processing device, electronic equipment and a computer readable storage medium.

Background

The three-dimensional high-precision map is a precondition for realizing automatic driving and auxiliary driving of the vehicle, and provides a main basis for accurate positioning and correct decision making of the automatic driving vehicle. The processing of the point cloud is a key link in the production of the three-dimensional high-precision map, and it can be understood that the point cloud refers to a point data set of the appearance surface of a product obtained by a measuring instrument in reverse engineering, and the point cloud can also be called as point cloud data.

In the prior art, some quality problems are likely to exist in the point cloud used in the three-dimensional high-precision map production, and once the point cloud with the quality problems flows into the map production operation stage, the point cloud-dependent algorithm is interfered, for example, the lane line modeling algorithm is interfered, so that the quality of the finally generated map data is reduced.

Disclosure of Invention

The embodiment of the invention provides a data processing method, a data processing device, electronic equipment and a computer readable storage medium, which are used for solving the problem that the quality of finally generated map data is reduced due to the fact that point clouds with quality problems are used when three-dimensional high-precision map production is carried out in the prior art.

In order to solve the technical problems, the invention is realized as follows:

In a first aspect, an embodiment of the present invention provides a data processing method, where the method includes:

according to the original point cloud, performing splicing treatment and fusion treatment to obtain a first treatment point cloud;

receiving input operation of a user under the condition that the first processing point cloud does not meet a first preset condition; the input operation is used for designating homonymous feature points in the first processing point cloud;

according to the homonymous feature points, carrying out fusion processing on the first processing point cloud to obtain a second processing point cloud;

and under the condition that the second processing point cloud meets the first preset condition and the distance between the same-name characteristic points in the second processing point cloud is smaller than the first preset distance, executing map production operation according to the second processing point cloud.

In a second aspect, an embodiment of the present invention provides a data processing apparatus, the apparatus including:

the first processing module is used for performing splicing processing and fusion processing according to the original point cloud to obtain a first processing point cloud;

the receiving module is used for receiving input operation of a user under the condition that the first processing point cloud does not meet a first preset condition; the input operation is used for designating homonymous feature points in the first processing point cloud;

The second processing module is used for carrying out fusion processing on the first processing point cloud according to the homonymous feature points to obtain a second processing point cloud;

and the execution module is used for executing map production operation according to the second processing point cloud when the second processing point cloud meets the first preset condition and the distance between the same-name characteristic points in the second processing point cloud is smaller than the first preset distance.

In a third aspect, an embodiment of the present invention provides an electronic device, including a processor, a memory, and a computer program stored in the memory and executable on the processor, the computer program implementing the steps of the data processing method described above when executed by the processor.

In a fourth aspect, embodiments of the present invention provide a computer readable storage medium having a computer program stored thereon, which when executed by a processor, implements the steps of the data processing method described above.

In the embodiment of the invention, in order to produce the three-dimensional high-precision map, the first processing point cloud can be obtained by performing splicing processing and fusion processing according to the original point cloud, and then whether the quality of the first processing point cloud meets the requirement can be judged according to whether the first processing point cloud meets the first preset condition. Under the condition that the quality of the first processing point cloud does not meet the requirement, the same-name characteristic points of the first processing point cloud can be appointed by manual intervention so as to perform fusion processing on the first processing point cloud according to the same-name characteristic points, and therefore a second processing point cloud with better quality is obtained. And then, judging whether the quality of the second processing point cloud meets the requirement according to whether the second processing point cloud meets the first preset condition and whether the distance of the homonymous characteristic points in the second processing point cloud is smaller than the first preset distance. And under the condition that the quality of the second processing point cloud meets the requirement, the map production operation can be executed according to the second processing point cloud. Therefore, compared with the prior art, the method and the device can effectively ensure the quality of the generated map data.

Drawings

In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings that are needed in the description of the embodiments of the present invention will be briefly described below, and it is obvious that the drawings in the following description are only some embodiments of the present invention, and other drawings may be obtained according to these drawings without inventive effort for a person skilled in the art.

FIG. 1 is one of the flowcharts of a data processing method provided by an embodiment of the present invention;

FIG. 2 is a schematic diagram of determining a layering rate of a first processing point cloud in an embodiment of the invention;

FIG. 3 is a second flowchart of a data processing method according to an embodiment of the present invention;

FIG. 4 is a block diagram of a data processing apparatus according to an embodiment of the present invention;

fig. 5 is a schematic structural diagram of an electronic device according to an embodiment of the present invention.

Detailed Description

The following description of the embodiments of the present invention will be made clearly and fully with reference to the accompanying drawings, in which it is evident that the embodiments described are some, but not all embodiments of the invention. All other embodiments, which can be made by those skilled in the art based on the embodiments of the invention without making any inventive effort, are intended to be within the scope of the invention.

The following first describes a data processing method provided by an embodiment of the present invention.

It should be noted that the data processing method provided by the embodiment of the invention is applied to electronic equipment. In particular, the electronic device may be a server or other device having data computing and processing functions. For easy understanding, the embodiment of the invention is described by taking the case that the electronic device is a server as an example.

Referring to fig. 1, a flowchart of a data processing method provided by an embodiment of the present invention is shown. As shown in fig. 1, the method comprises the steps of:

and step 101, performing splicing and fusion processing according to the original point cloud to obtain a first processing point cloud.

Prior to step 101, the server needs to obtain an original point cloud. Specifically, the collection device of the original point cloud (for convenience of explanation, the collection device of the original point cloud will be simply referred to as collection device hereinafter) may be deployed on a vehicle, and the server may download the original point cloud collected by the collection device.

Generally, when three-dimensional high-precision map generation is performed, the data volume of the original point cloud to be used is very huge, and even if the data volume of one city is subjected to thinning treatment, the data volume of one city can reach the TB level; here, TB is a unit of hard disk capacity, and 1 tb=1000 gb=1000000 MB. Therefore, when the original point cloud is downloaded, a plurality of threads can be called to perform batch asynchronous downloading. It can be understood that the point cloud may be divided into three levels of organizations, i.e. tile (i.e. 10 km by 10 km), node (i.e. 50 m by 50 m), cell (i.e. 1 m by 1 m), and when each downloading request adopts batch asynchronous downloading, data of multiple nodes can be obtained by downloading the request at a time. In addition, the acquisition device may employ an asynchronous response mode in order not to block the program.

Because the original point cloud is obtained by the downloading of the server, the server can perform splicing processing and fusion processing according to the original point cloud to obtain the first processing point cloud. The specific implementation forms of the splicing process and the fusion process are varied according to the original point cloud, and are described below by way of example.

In one implementation, according to the original point cloud, a splicing process and a fusion process are performed, including:

generating an intermediate point cloud according to the original point cloud; the intermediate point cloud comprises key information in an original point cloud;

and performing splicing treatment and fusion treatment on the intermediate point cloud.

In this implementation, for data of multiple nodes, the garbage information therein may be discarded, for example, the rgb (where r represents red, g represents green, and b represents blue) values set only for display are discarded, and key information that can be truly used for map production operation is retained, which may constitute an intermediate point cloud. Because the memory of the server is limited, after the intermediate point cloud is obtained, the intermediate point cloud can be written into the cache, and an index is constructed so as to facilitate the subsequent search of the point cloud.

When the three-dimensional high-precision map is required to be produced, the server can acquire the intermediate point cloud from the cache, and perform splicing and fusion processing on the intermediate point cloud to obtain a first processing point cloud. It can be appreciated that the fusion process of the point cloud functions to: noise, layering and redundancy of a multi-view point cloud overlapping area caused by measurement errors, matching errors and the like are eliminated, so that a single-layer point cloud model with clear detail characteristics and smooth surface is built.

Therefore, in the implementation form, the electronic device only needs to perform splicing processing and fusion processing on the intermediate point cloud comprising the key information in the original point cloud, so that the operation amount of the processing process can be effectively reduced, and the processing speed and the processing efficiency can be improved.

Of course, the implementation form of the splicing process and the fusion process according to the original point cloud is not limited to the above case. For example, after downloading the original point cloud acquired by the acquisition device, the server may directly write the original point cloud into the cache and construct an index; when the three-dimensional high-precision map is required to be produced, the server can directly acquire the original point cloud from the cache, and perform splicing and fusion processing on the original point cloud to obtain a first processing point cloud.

Step 102, receiving input operation of a user under the condition that the first processing point cloud does not meet a first preset condition; the input operation is used for designating the same-name feature points in the first processing point cloud.

Here, the first preset condition may be a basis for evaluating whether the quality of the point cloud is good or bad.

Specifically, if the first processing point cloud meets the first preset condition, the quality of the first processing point cloud may be considered to be better, and the subsequent map generation operation may be directly executed according to the first processing point cloud.

If the first processing point cloud does not meet the first preset condition, the quality of the first processing point cloud can be considered to be bad, and in order to avoid influencing the quality of map data generated later, the server can output prompt information for representing that the quality of the first processing point cloud is problematic in a voice, text and other modes, and input operation executed by a user according to the prompt information is received; the input operation is used for designating the same-name feature points in the first processing point cloud. In general, there are a huge number of homonymous feature points in the first processing point cloud, and the input operation may only specify part of homonymous feature points in the first processing point cloud, for example, a homonymous feature point on a specified board, a homonymous feature point on a rod, a homonymous feature point on the ground, and the like.

And 103, carrying out fusion processing on the first processing point cloud according to the homonymous feature points to obtain a second processing point cloud.

In step 103, further fusion processing may be performed on the first processing point cloud according to the homonymous feature points in the first processing point cloud specified by the input operation, so as to further eliminate noise, layering and redundancy of the multi-view point cloud area caused by measurement errors, matching errors and the like, so as to obtain a second processing point cloud with better quality compared with the first processing point cloud.

Step 104, executing map production operation according to the second processing point cloud when the second processing point cloud meets the first preset condition and the distance between the same-name feature points in the second processing point cloud is smaller than the first preset distance.

After the second processing point cloud is obtained, the electronic device may determine whether the second processing point cloud meets a first preset condition, and determine whether a distance between homonymous feature points in the second processing point cloud is smaller than a first preset distance. Here, the first preset condition is used as one basis for judging whether the quality of the point cloud is good or bad, and the same-name characteristic point distance is used as another basis for judging whether the quality of the point cloud is good or bad.

If the second processing point cloud meets the first preset condition and the distance between the same-name feature points in the second processing point cloud is smaller than the first preset distance, the quality of the second processing point cloud is good according to the two criteria for judging the quality of the point cloud, and at this time, the server can execute subsequent map generation operation according to the second processing point cloud, so that the quality of the subsequently generated map data can be effectively ensured.

If the second processing point cloud does not meet the first preset condition and/or the distance between the same-name feature points in the second processing point cloud is greater than or equal to the first preset distance, which indicates that the quality of the second processing point cloud is problematic according to at least one of the two criteria for judging whether the quality of the point cloud is good or bad, at this time, the server can output prompt information for representing that the quality of the second processing point cloud is problematic in a voice, text and other modes, and receive input operation executed by a user according to the prompt information; the input operation is used for designating the same-name feature points in the second processing point cloud. Then, the server may perform fusion processing on the second processing point cloud according to the same-name feature points specified by the input operation, and the subsequent process may refer to the description of the steps performed after the fusion processing is performed on the first processing point cloud, which is not described herein.

Optionally, the original point cloud includes point clouds acquired at different acquisition times, and the first processing point cloud includes processing point clouds corresponding to different acquisition times;

in the case that the first processing point cloud does not meet the first preset condition, before receiving the input operation of the user, the method further includes:

according to the processing point clouds corresponding to different acquisition times, the spatial position of the same object is respectively determined;

and determining whether the first processing point cloud meets a first preset condition according to the determined distance of each spatial position.

It should be noted that the object involved in this embodiment may be a card, a rod, or any object that is stationary in a real three-dimensional space.

In specific implementation, the acquisition device can periodically acquire the point cloud in the running process of the vehicle. Assuming that the point cloud collected in the period T1 is D1, the point cloud collected in the period T2 is D2, the point cloud collected in the period T3 is D3, and the point cloud collected in the period T4 is D4, D1, D2, D3, and D4 may be included in the original point cloud, and accordingly, D1 'corresponding to T1, D2' corresponding to T2, D3 'corresponding to T3, and D4' corresponding to T4 may be included in the first processing point cloud.

Next, the spatial position of the same rod (e.g., rod G) may be determined based on D1 'through D4', respectively. Assuming that the spatial position determined according to D1 'is W1, the spatial position determined according to D2' is W2, the spatial position determined according to D3 'is W3, and the spatial position determined according to D4' is W4, it may be determined whether the first processing point cloud satisfies the first preset condition according to the distance between two of the four W1 to W4.

Since the spatial position of G is fixed in the actual three-dimensional space, if the distances between W1 and W4 are very large, the quality problem of the first processing point cloud can be considered, and at this time, it can be determined that the first processing point cloud does not meet the first preset condition; in contrast, if the distances between W1 to W4 are very small, it may be determined that the first processing point cloud satisfies the first preset condition.

Therefore, in this embodiment, through determination of the spatial location distance, it can be determined more conveniently and reliably whether the first processing point cloud meets the first preset condition.

Optionally, before receiving the input operation of the user, the method further includes:

determining target information of a first processing point cloud; wherein the target information includes at least one of reflection information, layering information, dislocation information, and shielding information;

and determining whether the first processing point cloud meets a first preset condition according to the target information.

Here, the reflection information may include reflectivity; the layering information may include layering rate; the misalignment information may include a misalignment rate; the occlusion information may include an occlusion rate. From the target information, at least one of the following may be determined: (1) whether the reflectivity of the first processing point cloud meets the requirement; (2) whether layering in the first processing point cloud is severe; (3) whether the misalignment in the first processing point cloud is severe; (4) whether the occlusion situation in the first processing point cloud is serious.

And then, judging the quality of the first processing point cloud according to the judging result. Specifically, if the reflectivity of the first processing point cloud meets the requirement and the layering condition, the dislocation condition and the shielding condition are not serious, the quality of the first processing point cloud can be considered to be good, and then it can be determined that the first processing point cloud meets the first preset condition. If the reflectivity of the first processing point cloud does not meet the requirement and/or at least one of the layering condition, the dislocation condition and the shielding condition is serious, the quality problem of the first processing point cloud can be considered, and then it can be determined that the first processing point cloud does not meet the first preset condition.

Therefore, in this embodiment, according to the target information, it can be determined more conveniently and reliably whether the first processing point cloud meets the first preset condition.

Optionally, the collection device of the original point cloud is deployed on the vehicle;

determining target information of a first processing point cloud, including:

obtaining first vehicle track data; wherein the first vehicle track data corresponds to a first vehicle track;

selecting a plurality of checkpoints on corresponding tangents to each of the M track points on the first vehicle track; wherein M is an integer greater than or equal to 1, each check point corresponds to a ray, the starting point of the ray corresponding to any check point is the check point and the direction is the gravity direction;

Determining a target point cloud which is positioned in a stereoscopic region corresponding to the check point in the first processing point cloud aiming at each check point; the three-dimensional area corresponding to any check point is a cylindrical area with the ray corresponding to the check point as a central axis and the bottom radius as a preset radius;

and determining the layering trust of the first processing point cloud according to the target point cloud corresponding to each check point.

Here, the value of M may be 1, 2, 3, 4 or 5, the number of checkpoints selected on the tangent line corresponding to each track point may be 2, 3, 4, 5 or 6, and the preset radius may be 0.1 meter, 0.12 meter or 0.15 meter, and of course, the value of M, the number of checkpoints selected on each tangent line, and the value of the preset radius are not limited thereto, and may be specifically determined according to practical situations, and are not listed herein. In addition, the direction of gravity can be considered as the negative Z-axis direction.

In this embodiment, during the running of the vehicle, the vehicle may invoke a global positioning system (Global Positioning System, GPS) to perform positioning to obtain first vehicle track data corresponding to the first vehicle track. The vehicle may send the obtained first vehicle track data to the server, so that the server obtains the first vehicle track according to the first vehicle track data.

Assuming that the first vehicle track is track 200 in fig. 2, for the 3 track points J10, J20, and J30 on track 200, tangent lines of track 200 may be respectively made to obtain tangent line Q1, tangent line Q2, and tangent line Q3; wherein Q1 corresponds to J10, Q2 corresponds to J20, and Q3 corresponds to J30. Next, a plurality of checkpoints may be selected on Q1, Q2, and Q3, respectively, and the pitch of adjacent checkpoints thereon may be the same for any of Q1, Q2, and Q3, which may be 0.4 meters, 0.5 meters, 0.6 meters, and the like.

Specifically, assuming that for J10, five checkpoints are selected, where J10 is taken as a central checkpoint, and J11, J12, J13, and J14 are uniformly distributed on two sides of J10, for J11, it may be determined that, in the first processing point cloud, a target point cloud D11 located in a stereoscopic region corresponding to J11; for J12, determining a target point cloud D12 positioned in a stereoscopic region corresponding to J12 in the first processing point cloud; for J10, determining a target point cloud D10 in a stereoscopic region corresponding to J10 in the first processing point cloud; for J13, determining a target point cloud D13 positioned in a stereoscopic region corresponding to J13 in the first processing point cloud; for J14, a target point cloud D14 located in the stereoscopic region corresponding to J14 from among the first processing point clouds may be determined. The method for determining the target point cloud corresponding to the other checkpoints is described above, and is not described herein.

Then, the layering rate of the first processing point cloud can be determined according to the target point cloud corresponding to each check point, and a specific implementation form of determining the layering rate of the first processing point cloud is described below by way of example.

In one implementation, determining a layering rate of the first processing point cloud according to the target point cloud corresponding to each checkpoint includes:

for each check point, determining the point cloud which is positioned on the ground in the corresponding target point cloud; and determining the layering rate of the first processing point cloud according to the point clouds determined for each check point.

Here, the server may perform semantic segmentation of the point cloud to identify a point cloud located on the ground of the target point cloud according to the semantic segmentation result. Thereafter, a layering rate of the first processed point cloud may be determined based solely on the point cloud located at the ground. It should be noted that, according to the semantic segmentation result, the point cloud located at the tunnel or the intersection can be identified.

In the implementation form, the layering rate can be determined by only referencing the point cloud positioned on the ground, so that the accuracy and the efficiency of a determination result can be ensured.

In another implementation form, determining the layering rate of the first processing point cloud according to the target point cloud corresponding to each checkpoint includes:

For each check point, calculating the average value of the maximum height value and the rest height values except the minimum height value in the corresponding target point cloud, comparing the calculated average value with a preset average value, and determining whether layering occurs in the target point cloud corresponding to the check point according to the obtained comparison result; and determining the layering rate of the first processing point cloud according to whether layering occurs in the target point cloud corresponding to each check point.

Continuing with the example of fig. 2, after determining, for J10, a target point cloud D10 located in the stereoscopic region corresponding to J10 in the first processing point cloud, the height values (which may also be considered as coordinate values in the Z direction) in D10 may be sorted in order of magnitude, to obtain a sorted sequence of Z coordinates. Then, the maximum coordinate value and the minimum coordinate value in the ordered sequence can be discarded, the average value of other coordinate values in the ordered sequence is calculated, and the calculated average value is compared with a preset average value. If the calculated average value is larger than the preset average value, the D10 can be considered to be layered; otherwise, D10 may be considered to be not delaminated.

It should be noted that, by using the above-described strategy, it is also possible to determine whether or not delamination occurs in D11 to D14, respectively. For the checkpoints on the tangent line Q1 corresponding to J10, if the point cloud corresponding to the checkpoints exceeding a certain proportion (for example, 50%, 60%, etc.) is layered, it can be considered that the point cloud layering occurs at J10.

By adopting the mode, whether the point cloud layering occurs at the J20 and the J30 can be further determined. Then, the layering rate of the first processing point cloud can be determined according to the proportion of the track points where the point cloud layering occurs. Specifically, when the proportion of the trajectory points at which the point cloud layering occurs is 2%, it can be determined that the layering rate of the first processing point cloud is 2%.

It should be noted that if the layering rate of the first processing point cloud exceeds 2%, that is, if layering exists in more than 2% of the point clouds in the first processing point cloud, the first processing point cloud may be considered as not meeting the first preset condition.

In the implementation form, the layering rate can be conveniently and reliably determined by combining the Z coordinate value, and the maximum height value and the minimum height value are removed in the determination process, so that the error of the determination result can be effectively reduced.

Therefore, in this embodiment, the layering rate information of the first processing point cloud can be very conveniently and reliably determined by combining the vehicle track information.

before the splicing treatment and the fusion treatment are carried out according to the original point cloud, the method further comprises the following steps:

obtaining second vehicle track data; wherein the second vehicle track data corresponds to a second vehicle track;

Performing a first check on the second vehicle track data; wherein the first check includes at least one of an integrity check, a correctness check, and a precision check;

according to the original point cloud, performing splicing and fusion processing, including:

and under the condition that the first verification of the second vehicle track data is passed, performing splicing processing and fusion processing according to the original point cloud.

The method for acquiring the second vehicle track data may refer to the description of the method for acquiring the first vehicle track data, which is not described herein.

In general, the coordinate information in the original point cloud is determined by taking the track point in the second vehicle track as the origin, and under the condition that the first verification of the second vehicle track data is passed, the second vehicle track data is complete, accurate and reliable, and on the basis, the quality of the map data generated subsequently can be better ensured by performing the splicing processing and the fusion processing according to the original point cloud.

Optionally, the first verification comprises an integrity verification;

performing a first check on the second vehicle track data, including:

under the condition that the second vehicle track data corresponds to a plurality of second vehicle tracks, and vehicle tracks matched with each track in a preset map exist in the plurality of second vehicle tracks respectively, determining that the integrity verification of the second vehicle track data passes; otherwise, it is determined that the integrity check on the second vehicle track data is not passed.

Here, the preset map may be a two-dimensional map; any track in the preset map and the track of the vehicle matched with the track in the preset map can be considered to be the same track, a matching algorithm can be adopted when the track is matched, and the matching algorithm can adopt a hidden Markov model.

In general, when a high-precision map is generated, an original point cloud needs to be acquired on each road in a preset map, and when the original point cloud is acquired, a vehicle can call a GPS at the same time to acquire track data. In this way, when the second vehicle track data corresponds to a plurality of second vehicle tracks and each track in the preset map exists in the plurality of second vehicle tracks, the second vehicle track data can be considered to be complete, and the integrity check of the second vehicle track data is passed; otherwise, it is determined that its integrity check is not passed.

Therefore, in this embodiment, based on the preset map, it may be more convenient and reliable to determine whether the integrity check of the second vehicle track data passes.

Optionally, the first check comprises a correctness check;

performing a first check on the second vehicle track data, including:

determining that the correctness checking pass of the second vehicle track data is determined under the condition that the second vehicle track data meets a second preset condition, the proportion that the distance between adjacent track points is smaller than the second preset distance in the second vehicle track is larger than the first preset proportion and the proportion that the distance between adjacent track points is smaller than the third preset distance is equal to the second preset proportion according to the second vehicle track data; otherwise, determining that the correctness check of the second vehicle track data is not passed;

The third preset distance is greater than the second preset distance, and the second preset proportion is greater than the first preset proportion.

Here, the second vehicle track data may exist in the form of a file, which may also be referred to as EOUT. The second preset distance may be 5 meters, the first preset ratio may be 99 meters, the third preset distance may be 10 meters, and the second preset ratio may be 100%, however, the values of the second preset distance, the first preset ratio, the third preset distance and the second preset ratio are not limited thereto, and may be specifically determined according to practical situations, and are not listed here.

In specific implementation, the correctness checking may refer to the following information:

(1) EOUT is correctly formatted (i.e., EOUT is correctly resolved);

(2) The time stamp of EOUT increases sequentially in units of 0.05;

(3) The difference between the start-stop end time of EOUT and the start-stop end time of GPS is less than 5 seconds;

(4) The starting and ending time stamp of the EOUT comprises all time stamps of the point cloud and the picture, namely, when the acquisition equipment acquires the point cloud, the GPS also records position information when the camera shoots an image so as to obtain corresponding vehicle track data;

(5) The plane distance between two adjacent points (i.e. the distance between two adjacent track points) is less than 5 m and more than or equal to 99%, and less than 10 m and equal to 100%;

(6) The track elevation difference of the 0.1 second time interval is more than or equal to 99 percent with less than 0.25 meter.

It should be noted that, when the above (1) to (3) are all satisfied, it may be considered that the second vehicle track data satisfies the second preset condition, and on the basis of this, if the above (4) to (6) are also all satisfied, it may be determined that the correctness check of the second vehicle track data passes. If any one of the above (1) to (6) is not satisfied, it may be determined that the correctness check of the second vehicle track data is not passed.

Therefore, in this embodiment, by adopting the above manner, it can be determined more conveniently and reliably whether the correctness check of the second vehicle track data passes.

Optionally, the first verification includes a precision verification;

performing a first check on the second vehicle track data, including:

determining that the accuracy verification of the second vehicle track data passes under the condition that the track jump value of the second vehicle track is smaller than a preset jump value and the vehicle coordinate difference value of the second vehicle track and the third vehicle track corresponding to the same space position is smaller than a preset difference value according to the second vehicle track data; otherwise, determining that the accuracy verification of the second vehicle track data is not passed;

The third vehicle track is a track matched with the second vehicle track in a preset map.

Here, the preset jump value may be 15 cm, 20 cm or 25 cm, and of course, the value of the preset jump value is not limited to this, and may be specifically determined according to the actual situation, which is not limited in this embodiment.

Here, the preset map may be a two-dimensional map, and in addition, the third vehicle track and the second vehicle track may be considered as the same track, a matching algorithm may be adopted when matching the tracks, and the matching algorithm may be a hidden markov model.

In specific implementation, the accuracy check may refer to the following information:

(1) The height difference of different tracks on the same road is less than 2 meters;

(2) The level difference of different tracks on the same road is less than 2 meters;

(3) The single trip trace point jump is less than 20 cm.

Note that, when the above-described (1) to (3) are all satisfied, it can be considered that the accuracy check of the second vehicle track data passes; in the case where at least one of the above-described (1) to (3) is not satisfied, it can be considered that the accuracy check of the second vehicle track data is not passed.

Therefore, in this embodiment, by adopting the above manner, it can be determined more conveniently and reliably whether the accuracy check of the second vehicle track data passes.

Optionally, before the splicing process and the fusion process, the method further includes:

performing second check on the original point cloud; wherein the second check comprises at least one of an integrity check and a precision check;

and under the condition that the second check on the original point cloud passes, performing splicing processing and fusion processing according to the original point cloud.

Under the condition that the second check of the original point cloud is passed, the original point cloud is complete and accurate data, and on the basis, the quality of map data generated subsequently can be better ensured by performing splicing and fusion processing according to the original point cloud.

obtaining a three-dimensional bounding box;

determining a storage path corresponding to the three-dimensional bounding box;

and acquiring an original point cloud according to the storage path.

After the original point cloud acquired by the acquisition device is downloaded, the server may write the original point cloud into the cache, and construct an index, where the index may include a correspondence between a three-dimensional bounding box of the original point cloud and a storage path. Therefore, the corresponding storage path can be determined according to the index only by knowing the three-dimensional bounding box, and the original point cloud is acquired from the storage path so as to facilitate subsequent processing. It can be seen that, in this embodiment, the operation of obtaining the original point cloud is very convenient to implement.

The following describes the implementation of the present embodiment in detail with reference to fig. 3, by way of a specific example.

As shown in fig. 3, track profile data may be acquired first and a track profile admission decision may be made. Here, the track data corresponds to the second vehicle track data above, and the track data admission determination corresponds to the first verification of the second vehicle track data above, so that the track data admission determination can check the integrity of the cloud acquisition track and the accuracy of the cloud acquisition track.

If the track material admission decision does not pass, the track material data may be re-collected. And if the admission judgment of the track data is passed, the admission judgment of the original data of the point cloud can be carried out. Here, the point cloud raw data admission determination corresponds to the above first verification of the raw point cloud, and then the point cloud raw data admission determination can check the integrity of the raw point cloud and check the accuracy of the raw point cloud.

If the point cloud original data admission judgment does not pass, the original point cloud can be collected again. If the point cloud original data admission judgment passes, the point cloud splicing and the point cloud automatic fusion can be sequentially carried out, and the first processing point cloud can be obtained through the point cloud splicing and the point cloud automatic fusion.

Next, a point cloud auto-fusion accurate determination may be made. Here, the point cloud automatic fusion accurate determination is equivalent to the above determination as to whether the first processing point cloud satisfies the first preset condition, and then the point cloud automatic fusion accurate determination can check: whether the reflectivity of the point cloud data meets the requirement, whether the point cloud data is layered, whether the point cloud data is misplaced, and whether the point cloud data is shielded.

If the point cloud automatic fusion accurate determination passes, map production operation can be executed according to the first processing point cloud, for example, lane line modeling is directly carried out. If the automatic point cloud fusion accuracy judgment is not passed, the point cloud characteristic points can be manually adjusted, and the point cloud artificial fusion is carried out, which is equivalent to the fusion processing of the first processing point cloud according to the same-name characteristic points in the first processing point cloud appointed by the input operation, so as to obtain the second processing point cloud.

Next, a point cloud artificial fusion accuracy determination may be performed. Here, the point cloud artificial fusion accurate determination is equivalent to the above determination whether the second processing point cloud satisfies the first preset condition, and whether the distance of the homonymous feature points in the second processing point cloud is smaller than the first preset distance.

If the point cloud manual fusion accurate determination is not passed, the point cloud characteristic points can be continuously manually adjusted. If the point cloud manual fusion accurate determination passes, map production operation can be executed according to the second processing point cloud, for example, lane line modeling is directly carried out.

As can be seen from the above examples, in the present embodiment, a systematic point cloud quality check may be performed to ensure that the quality of the point cloud for performing the map production job operation meets the requirements. The inspection process is specifically divided into three parts, namely: (1) The method mainly comprises the steps of checking original data, namely checking the integrity and the accuracy of a point cloud acquisition track/point cloud original file; the point cloud acquisition track is second vehicle track data, and the point cloud original data is original point cloud; (2) Checking the automatic fused point cloud, namely checking whether layering/error/shielding and other conditions exist in the point cloud, whether characteristics such as cards/rods extracted in the fusion process are accurate and whether the fused distance is larger than a threshold value; (3) checking the point cloud after artificial fusion: mainly, whether layering/error/shielding and other conditions exist in the check point cloud, and whether the manually specified distance between the same name points after fusion is larger than a first preset distance or not.

Experiments show that the quality of the point cloud is checked by adopting the strategy, the data quality after the point cloud fusion can be effectively controlled, the layered/misplaced/shielded point cloud is ensured not to flow to the subsequent operation links, the problematic point cloud can be reduced from 10% to 1%, the labor cost consumed for checking the quality of the point cloud is reduced from 5 persons to 1 person, and the automatic turning-out time of each tile is controlled within 30 minutes.

It should be noted that, because the data size of the point cloud is often very large, when the point cloud is inspected, in order to increase the inspection speed, the point cloud can be uniformly divided into a plurality of parts, and a plurality of threads (the number of specific threads is related to the number of machine cores) are called up, and each thread inspects the data quality of the point cloud corresponding to a track. In the checking process, the point cloud files are swapped in and out, in order to ensure enough memory, the maximum number of files and the minimum number of files in the memory are required to be set, wherein a first-in first-out (First Input First Output, FIFO) strategy can be adopted, when the maximum number of files is exceeded, the earliest opened files which are not used currently are required to be closed until the number of files is equal to the minimum number of files, and the memory occupation is kept in dynamic balance in this way

In summary, compared with the prior art, the present embodiment can effectively ensure the quality of the generated map data.

The following describes a data processing apparatus provided in an embodiment of the present invention.

Referring to FIG. 4, a block diagram of a data processing apparatus 400 according to an embodiment of the present invention is shown. As shown in fig. 4, the data processing apparatus 400 includes:

the first processing module 401 is configured to perform a splicing process and a fusion process according to the original point cloud, so as to obtain a first processing point cloud;

a receiving module 402, configured to receive an input operation of a user if the first processing point cloud does not meet a first preset condition; the input operation is used for designating homonymous feature points in the first processing point cloud;

the second processing module 403 is configured to perform fusion processing on the first processing point cloud according to the homonymous feature points, so as to obtain a second processing point cloud;

and the execution module 404 is configured to execute the map production operation according to the second processing point cloud when the second processing point cloud meets the first preset condition and the distance between the same-name feature points in the second processing point cloud is smaller than the first preset distance.

The data processing apparatus 400 further includes:

the first determining module is used for respectively determining the spatial position of the same object according to the processing point clouds corresponding to different acquisition times before receiving the input operation of the user under the condition that the first processing point clouds do not meet the first preset condition;

and the second determining module is used for determining whether the first processing point cloud meets a first preset condition according to the determined distance of each spatial position.

Optionally, the data processing apparatus 400 further comprises:

the third determining module is used for determining target information of the first processing point cloud before receiving input operation of a user under the condition that the first processing point cloud does not meet a first preset condition; wherein the target information includes at least one of reflection information, layering information, dislocation information, and shielding information;

and the fourth determining module is used for determining whether the first processing point cloud meets a first preset condition according to the target information.

a third determination module, comprising:

an obtaining sub-module for obtaining first vehicle track data; wherein the first vehicle track data corresponds to a first vehicle track;

a selection sub-module, configured to select, for each of M track points on the first vehicle track, a plurality of checkpoints on a tangent line corresponding to the track point; wherein M is an integer greater than or equal to 1, each check point corresponds to a ray, the starting point of the ray corresponding to any check point is the check point and the direction is the gravity direction;

The first determining submodule is used for determining target point clouds which are located in the stereoscopic region corresponding to the check point in the first processing point clouds aiming at each check point; the three-dimensional area corresponding to any check point is a cylindrical area with the ray corresponding to the check point as a central axis and the bottom radius as a preset radius;

and the second determining submodule is used for determining layering information of the first processing point cloud according to the target point cloud corresponding to each check point.

Optionally, the second determining submodule is specifically configured to:

for each check point, determining the point cloud which is positioned on the ground in the corresponding target point cloud; determining layering information of the first processing point cloud according to the point clouds determined for each check point;

optionally, the second determining submodule is specifically configured to:

Optionally, the first processing module 401 includes:

the generation sub-module is used for generating an intermediate point cloud according to the original point cloud; the intermediate point cloud comprises key information in an original point cloud;

and the processing sub-module is used for performing splicing processing and fusion processing on the intermediate point cloud.

the data processing apparatus 400 further includes:

the first acquisition module is used for acquiring second vehicle track data before performing splicing processing and fusion processing according to the original point cloud; wherein the second vehicle track data corresponds to a second vehicle track;

the first verification module is used for carrying out first verification on the second vehicle track data; wherein the first check includes at least one of an integrity check, a correctness check, and a precision check;

the first processing module 401 is specifically configured to:

Optionally, the first verification comprises an integrity verification;

the first verification module is specifically configured to:

Optionally, the first check comprises a correctness check;

the first verification module is specifically configured to:

Optionally, the first verification includes a precision verification;

the first verification module is specifically configured to:

Optionally, the data processing apparatus 400 further comprises:

the second obtaining module is used for obtaining a three-dimensional bounding box before the splicing treatment and the fusion treatment are carried out according to the original point cloud;

a fifth determining module, configured to determine a storage path corresponding to the three-dimensional bounding box;

and the third obtaining module is used for obtaining the original point cloud according to the storage path.

The electronic device provided by the embodiment of the invention is explained below.

Referring to fig. 5, a schematic structural diagram of an electronic device 500 according to an embodiment of the present invention is shown. As shown in fig. 5, the electronic device 500 includes: a processor 501, a memory 503, a user interface 504 and a bus interface.

A processor 501 for reading the program in the memory 503, performing the following procedures:

and under the condition that the second processing point cloud meets the first preset condition and the distance between the homonymous feature points in the second processing point cloud is smaller than the first preset distance, executing map production operation according to the second processing point cloud.

In fig. 5, a bus architecture may comprise any number of interconnected buses and bridges, with one or more processors, represented in particular by processor 501, and various circuits of memory, represented by memory 503, linked together. The bus architecture may also link together various other circuits such as peripheral devices, voltage regulators, power management circuits, etc., which are well known in the art and, therefore, will not be described further herein. The bus interface provides an interface. The user interface 504 may also be an interface capable of interfacing with an inscribed desired device for a different user device, including but not limited to a keypad, display, speaker, microphone, joystick, etc.

The processor 501 is responsible for managing the bus architecture and general processing, and the memory 503 may store data used by the processor 501 in performing operations.

the processor 501 is further configured to:

under the condition that the first processing point cloud does not meet the first preset condition, before receiving input operation of a user, respectively determining the spatial position of the same object according to the processing point clouds corresponding to different acquisition times;

Optionally, the processor 501 is further configured to:

determining target information of the first processing point cloud before receiving input operation of a user under the condition that the first processing point cloud does not meet a first preset condition; wherein the target information includes at least one of reflection information, layering information, dislocation information, and shielding information;

the processor 501 is specifically configured to:

and determining the layering rate of the first processing point cloud according to the target point cloud corresponding to each check point.

Optionally, the processor 501 is specifically configured to:

the processor 501 is specifically configured to:

Optionally, the first verification comprises an integrity verification;

the processor 501 is specifically configured to:

Optionally, the first check comprises a correctness check;

the processor 501 is specifically configured to:

Optionally, the first verification includes a precision verification;

the processor 501 is specifically configured to:

Preferably, the embodiment of the present invention further provides an electronic device, including a processor 501, a memory 503, and a computer program stored in the memory 503 and capable of running on the processor 501, where the computer program when executed by the processor 501 implements each process of the foregoing embodiment of the data processing method, and the same technical effects can be achieved, and for avoiding repetition, a description is omitted herein.

The embodiment of the invention also provides a computer readable storage medium, on which a computer program is stored, which when executed by a processor, implements the respective processes of the above-mentioned data processing method embodiment, and can achieve the same technical effects, and in order to avoid repetition, the description is omitted here. Wherein the computer readable storage medium is selected from Read-Only Memory (ROM), random access Memory (Random Access Memory, RAM), magnetic disk or optical disk.

The embodiments of the present invention have been described above with reference to the accompanying drawings, but the present invention is not limited to the above-described embodiments, which are merely illustrative and not restrictive, and many forms may be made by those having ordinary skill in the art without departing from the spirit of the present invention and the scope of the claims, which are to be protected by the present invention.

Claims

1. A method of data processing, the method comprising:

receiving input operation of a user under the condition that the first processing point cloud does not meet a first preset condition; the input operation is used for designating homonymous feature points in the first processing point cloud, the first preset condition is used for judging the quality of the point cloud, and the original point cloud is the point cloud acquired by acquisition equipment deployed on a vehicle;

2. The method of claim 1, wherein the original point cloud comprises point clouds acquired at different acquisition times, and the first processing point cloud comprises processing point clouds corresponding to different acquisition times;

the method further comprises, before receiving an input operation of a user in the case that the first processing point cloud does not meet a first preset condition:

3. The method of claim 1, wherein, in the case where the first processing point cloud does not satisfy the first preset condition, before receiving the input operation of the user, the method further comprises:

determining target information of the first processing point cloud; wherein the target information includes at least one of reflection information, layering information, misalignment information, and occlusion information;

4. The method of claim 3, wherein the collection device of the raw point cloud is deployed on a vehicle;

the determining the target information of the first processing point cloud includes:

selecting a plurality of checkpoints on corresponding tangents to each of M track points on the first vehicle track; wherein M is an integer greater than or equal to 1, each check point corresponds to a ray, the starting point of the ray corresponding to any check point is the check point and the direction is the gravity direction;

5. The method of claim 4, wherein determining the layering rate of the first processing point cloud from the target point cloud corresponding to each checkpoint comprises:

6. The method of claim 4, wherein determining the layering rate of the first processing point cloud from the target point cloud corresponding to each checkpoint comprises:

7. The method of claim 1, wherein the performing a stitching process and a fusion process according to the original point cloud comprises:

generating an intermediate point cloud according to the original point cloud; the intermediate point cloud comprises key information in the original point cloud;

8. The method of claim 1, wherein the collection device of the raw point cloud is deployed on a vehicle;

and performing splicing and fusion processing according to the original point cloud, wherein the method comprises the following steps:

9. The method of claim 8, wherein the first check comprises an integrity check;

The first checking the second vehicle track data includes:

determining that the integrity verification of the second vehicle track data passes under the condition that the second vehicle track data corresponds to a plurality of second vehicle tracks and the vehicle tracks matched with each track in a preset map exist in the plurality of second vehicle tracks respectively; otherwise, determining that the integrity check of the second vehicle track data is not passed.

10. The method of claim 8, wherein the first check comprises a correctness check;

the first checking the second vehicle track data includes:

determining that the correctness checking of the second vehicle track data passes under the condition that the second vehicle track data meets a second preset condition, and determining that the proportion of the adjacent track point distance smaller than the second preset distance in the second vehicle track is larger than the first preset proportion and the proportion of the adjacent track point distance smaller than the third preset distance is equal to the second preset proportion according to the second vehicle track data; otherwise, determining that the correctness check of the second vehicle track data is not passed;

11. The method of claim 8, wherein the first check comprises a precision check;

the first checking the second vehicle track data includes:

12. A data processing apparatus, the apparatus comprising:

the receiving module is used for receiving input operation of a user under the condition that the first processing point cloud does not meet a first preset condition; the input operation is used for designating homonymous feature points in the first processing point cloud, and the first preset condition is used for judging the quality of the point cloud;

the execution module is configured to execute a map production operation according to the second processing point cloud when the second processing point cloud meets the first preset condition and the distance between the same-name feature points in the second processing point cloud is smaller than the first preset distance, where the original point cloud is a point cloud collected by a collection device deployed on a vehicle.

13. The apparatus of claim 12, wherein the original point cloud comprises point clouds acquired at different acquisition times, and the first processing point cloud comprises processing point clouds corresponding to different acquisition times;

the apparatus further comprises:

the first determining module is used for determining the spatial position of the same object according to the processing point clouds corresponding to different acquisition times before receiving the input operation of the user under the condition that the first processing point cloud does not meet the first preset condition;

14. The apparatus of claim 12, wherein the apparatus further comprises:

a third determining module, configured to determine target information of the first processing point cloud before receiving an input operation of a user if the first processing point cloud does not meet a first preset condition; wherein the target information includes at least one of reflection information, layering information, misalignment information, and occlusion information;

15. The apparatus of claim 14, wherein the collection device of the raw point cloud is deployed on a vehicle;

the third determining module includes:

a selecting sub-module, configured to select, for each of M track points on the first vehicle track, a plurality of checkpoints on a tangent line corresponding to the track point; wherein M is an integer greater than or equal to 1, each check point corresponds to a ray, the starting point of the ray corresponding to any check point is the check point and the direction is the gravity direction;

The first determining submodule is used for determining target point clouds which are located in a stereoscopic region corresponding to the check point in the first processing point clouds aiming at each check point; the three-dimensional area corresponding to any check point is a cylindrical area with the ray corresponding to the check point as a central axis and the bottom radius as a preset radius;

and the second determining submodule is used for determining the layering rate of the first processing point cloud according to the target point cloud corresponding to each check point.

16. The apparatus according to claim 15, wherein the second determination submodule is specifically configured to:

17. The apparatus according to claim 15, wherein the second determination submodule is specifically configured to:

18. The apparatus of claim 12, wherein the collection device of the raw point cloud is deployed on a vehicle;

the apparatus further comprises:

the first acquisition module is used for acquiring second vehicle track data before the splicing treatment and the fusion treatment according to the original point cloud; wherein the second vehicle track data corresponds to a second vehicle track;

the first processing module is specifically configured to:

19. An electronic device comprising a processor, a memory, a computer program stored on the memory and executable on the processor, which computer program, when executed by the processor, implements the steps of the data processing method according to any one of claims 1 to 11.

20. A computer-readable storage medium, on which a computer program is stored, which computer program, when being executed by a processor, implements the steps of the data processing method according to any of claims 1 to 11.