CN113887351B - Obstacle detection method and obstacle detection device for unmanned driving - Google Patents
Obstacle detection method and obstacle detection device for unmanned driving Download PDFInfo
- Publication number
- CN113887351B CN113887351B CN202111104226.2A CN202111104226A CN113887351B CN 113887351 B CN113887351 B CN 113887351B CN 202111104226 A CN202111104226 A CN 202111104226A CN 113887351 B CN113887351 B CN 113887351B
- Authority
- CN
- China
- Prior art keywords
- point cloud
- cloud data
- obstacle
- range
- determining
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/10—Segmentation; Edge detection
- G06T7/13—Edge detection
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/50—Depth or shape recovery
- G06T7/521—Depth or shape recovery from laser ranging, e.g. using interferometry; from the projection of structured light
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/10—Image acquisition modality
- G06T2207/10028—Range image; Depth image; 3D point clouds
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Computer Vision & Pattern Recognition (AREA)
- General Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- Artificial Intelligence (AREA)
- Life Sciences & Earth Sciences (AREA)
- Optics & Photonics (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Bioinformatics & Computational Biology (AREA)
- Evolutionary Biology (AREA)
- Evolutionary Computation (AREA)
- General Engineering & Computer Science (AREA)
- Optical Radar Systems And Details Thereof (AREA)
- Traffic Control Systems (AREA)
Abstract
The specification discloses a method and a device for detecting an obstacle in unmanned driving, relates to the field of unmanned driving, and aims to obtain point cloud data acquired by unmanned driving equipment, determine a corresponding undetermined detection range of the obstacle in the point cloud data, determine the point cloud data in the undetermined detection range, serve as target point cloud data, and perform data enhancement on the target point cloud data based on standard point cloud data to obtain enhanced point cloud data. And then, determining point cloud characteristics corresponding to the enhanced point cloud data, performing range detection on the obstacle according to the point cloud characteristics to obtain an optimized detection range, and performing obstacle detection on the obstacle based on the optimized detection range, so that the enhanced point cloud data is densified, partial deletion is filled, and the range of the obstacle is determined more accurately through the enhanced point cloud data.
Description
Technical Field
The present disclosure relates to the field of unmanned driving, and more particularly, to an obstacle detection method and an obstacle detection apparatus for unmanned driving.
Background
In the unmanned technology, the obstacles around the unmanned equipment are accurately detected (such as the positions of the obstacles are determined, the types of the obstacles are determined, and the like), so that the unmanned equipment can accurately avoid the obstacles.
In the prior art, the unmanned device may determine a 3D range where an obstacle is located in point cloud data through collected point cloud data, so as to perform subsequent operations such as determining a type of the obstacle and determining a location where the obstacle is located according to the 3D range, and since the point cloud data is sparse and scanning of the obstacle is performed through a radar, due to a limitation of a bearing, some locations may not be scanned, so that some edge regions in the collected point cloud related to the obstacle may be missing, and therefore, an error may exist in the determined 3D range where the obstacle is located in the point cloud data.
Therefore, how to improve the accuracy of determining the 3D range of the obstacle in the point cloud data is an urgent problem to be solved.
Disclosure of Invention
The present specification provides an obstacle detection method and an obstacle detection apparatus for unmanned driving, which partially solve the above problems in the prior art.
The technical scheme adopted by the specification is as follows:
the present specification provides an obstacle detection method for unmanned driving, including:
acquiring point cloud data acquired by unmanned equipment, and determining a corresponding undetermined detection range of an obstacle in the point cloud data;
determining point cloud data in the undetermined detection range as target point cloud data;
performing data enhancement on the target point cloud data based on standard point cloud data and the undetermined detection range to obtain enhanced point cloud data, wherein the standard point cloud data is used for performing densification and missing filling on the target point cloud data;
and determining point cloud features corresponding to the enhanced point cloud data, performing range detection on the obstacle according to the point cloud features to obtain an optimized detection range, and performing obstacle detection on the obstacle based on the optimized detection range.
Optionally, based on the standard point cloud data and the undetermined detection range, performing data enhancement on the target point cloud data to obtain enhanced point cloud data, and specifically including:
adding the standard point cloud data to the target point cloud data;
and adjusting standard point cloud data outside the undetermined detection range, and/or adjusting point cloud data in the undetermined detection range after the standard point cloud data are superposed to obtain enhanced point cloud data.
Optionally, adjusting the standard point cloud data located outside the pending detection range specifically includes:
and adjusting the laser reflectivity of the standard point cloud data outside the undetermined detection range to be not more than the set reflectivity, wherein for any point cloud point, if the laser reflectivity of the point cloud point is higher, the point cloud point is more prominent in the point cloud.
Optionally, adjusting the point cloud data in the undetermined detection range after the standard point cloud data is superimposed, specifically including:
adjusting the point cloud data in the undetermined detection range after the standard point cloud data is superposed, wherein the point cloud data is the more prominent the point cloud point in the point cloud if the laser reflectivity of the point cloud point is higher for any point cloud point.
Optionally, determining the point cloud features corresponding to the enhanced point cloud data specifically includes:
determining an original coordinate corresponding to each cloud point in the enhanced point cloud data, wherein the original coordinate is a coordinate under a coordinate system taking acquisition equipment for acquiring the target point cloud data as a center;
determining a reference point in the enhanced point cloud data, and converting an original coordinate corresponding to each point cloud point of the enhanced point cloud data into a coordinate under a coordinate system with the reference point as a center to obtain a converted coordinate corresponding to the point cloud point;
and determining the point cloud characteristics corresponding to the enhanced point cloud data according to the converted coordinates corresponding to the cloud points of each point in the enhanced point cloud data.
Optionally, determining a corresponding undetermined detection range of the obstacle in the point cloud data specifically includes:
determining global features corresponding to the point cloud data, and determining a basic range of the obstacle in the point cloud data based on the global features;
determining partial features of the global features which belong to the basic range;
and determining the range to be detected according to the partial characteristics.
Optionally, determining the pending detection range according to the partial feature specifically includes:
determining each candidate frame corresponding to the obstacle and the confidence corresponding to each candidate frame according to the partial features;
and selecting a target frame from the candidate frames according to the candidate frames and the confidence degrees corresponding to the candidate frames, and determining the range to be detected according to the target frame.
Optionally, selecting a target frame from the candidate frames according to the candidate frames and the confidence degrees corresponding to the candidate frames specifically includes:
in the N round of screening, screening a candidate frame with the highest confidence coefficient from a candidate frame set corresponding to the N round as a candidate frame screened by the N round, and removing a candidate frame, of which the coincidence rate with the candidate frame with the highest confidence coefficient exceeds a set coincidence rate, from the candidate frame set corresponding to the N round to obtain a candidate frame set corresponding to the (N + 1) th round until a preset screening condition is met, and obtaining a remaining candidate frame, wherein for any round of screening, if a candidate frame, of which the coincidence rate with the candidate frame screened by the round exceeds the set coincidence rate, is not found from the candidate frame set corresponding to the round, the candidate frame set meeting the screening condition is determined, and the candidate frame set obtained after the round of screening is used as the remaining candidate frame, wherein N is a positive integer;
and determining the target frame according to the candidate frame screened out in each round and the residual candidate frame.
Optionally, point cloud data acquired by the unmanned device is acquired, and a corresponding undetermined detection range of the obstacle in the point cloud data is determined, and the method specifically includes:
inputting the point cloud data into a first network, and determining a corresponding undetermined detection range of the obstacle in the point cloud data;
determining point cloud characteristics corresponding to the enhanced point cloud data, and performing range detection on the obstacle according to the point cloud characteristics to obtain an optimized detection range, wherein the method specifically comprises the following steps:
and determining point cloud characteristics corresponding to the enhanced point cloud data through a second network, and performing range detection on the barrier according to the point cloud characteristics to obtain an optimized detection range, wherein the number of network layers of the second network is smaller than that of the first network.
Optionally, training the second network specifically includes:
acquiring a training sample, wherein the training sample comprises obstacle point cloud data and a labeling range corresponding to the obstacle point cloud data;
based on the standard point cloud data and the labeling range, enhancing the obstacle point cloud data to obtain enhanced point cloud data corresponding to the obstacle point cloud data;
inputting the enhanced point cloud data into a second network to be trained to obtain a prediction range;
and training the second network by taking the minimum deviation between the prediction range and the labeling range as a target.
The present specification provides an obstacle detection device for unmanned driving, including:
the system comprises an acquisition module, a detection module and a detection module, wherein the acquisition module is used for acquiring point cloud data acquired by unmanned equipment and determining a corresponding undetermined detection range of an obstacle in the point cloud data;
the determining module is used for determining point cloud data in the undetermined detection range as target point cloud data;
the enhancement module is used for carrying out data enhancement on the target point cloud data based on standard point cloud data to obtain enhanced point cloud data, and the standard point cloud data is used for carrying out densification on the target point cloud data;
and the optimization module is used for determining point cloud characteristics corresponding to the enhanced point cloud data, performing range detection on the obstacle according to the point cloud characteristics to obtain an optimized detection range, and performing obstacle detection on the obstacle based on the optimized detection range.
The present specification provides a computer-readable storage medium storing a computer program which, when executed by a processor, implements the above-described obstacle detection method for unmanned driving.
The present specification provides an unmanned device comprising a memory, a processor and a computer program stored on the memory and executable on the processor, the processor implementing the above described obstacle detection method for unmanned driving when executing the program.
The technical scheme adopted by the specification can achieve the following beneficial effects:
according to the method, the point cloud data acquired by the unmanned equipment are acquired, the undetermined detection range of the obstacle corresponding to the point cloud data is determined, the point cloud data in the undetermined detection range is determined and used as target point cloud data, and data enhancement is performed on the target point cloud data based on the standard point cloud data to obtain the enhanced point cloud data. And then determining point cloud characteristics corresponding to the enhanced point cloud data, performing range detection on the obstacle according to the point cloud characteristics to obtain an optimized detection range, and performing obstacle detection on the obstacle based on the optimized detection range.
According to the method, the standard point cloud data can be added into the target point cloud data corresponding to the obstacle based on the undetermined detection range, so that the enhanced point cloud data is densified, and partial missing positions (such as edges) are filled, so that the range of the obstacle can be more accurately determined through the enhanced point cloud data.
Drawings
The accompanying drawings, which are included to provide a further understanding of the specification and are incorporated in and constitute a part of this specification, illustrate embodiments of the specification and together with the description serve to explain the principles of the specification and not to limit the specification in a limiting sense. In the drawings:
fig. 1 is a schematic flow chart of an obstacle detection method for unmanned driving in the present specification;
FIG. 2 is a schematic diagram of enhanced point cloud data provided herein;
FIG. 3 is a schematic diagram of an obstacle detection apparatus for unmanned aerial vehicle provided herein;
fig. 4 is a schematic view of the drone corresponding to fig. 1 provided by the present description.
Detailed Description
In order to make the objects, technical solutions and advantages of the present disclosure more clear, the technical solutions of the present disclosure will be clearly and completely described below with reference to the specific embodiments of the present disclosure and the accompanying drawings. It is to be understood that the embodiments described are only a few embodiments of the present disclosure, and not all embodiments. All other embodiments obtained by a person of ordinary skill in the art based on the embodiments in the present specification without any creative effort belong to the protection scope of the present specification.
The technical solutions provided by the embodiments of the present description are described in detail below with reference to the accompanying drawings.
Fig. 1 is a schematic flow chart of an obstacle detection method for unmanned driving in this specification, which specifically includes the following steps:
s101: the method comprises the steps of obtaining point cloud data collected by unmanned equipment, and determining a corresponding undetermined detection range of an obstacle in the point cloud data.
S102: and determining the point cloud data in the undetermined detection range as target point cloud data.
S103: and based on the standard point cloud data and the undetermined detection range, performing data enhancement on the target point cloud data to obtain enhanced point cloud data, wherein the standard point cloud data is used for performing densification on the target point cloud data.
In the field of unmanned driving, unmanned equipment needs to detect obstacles and realize subsequent obstacle avoidance and other operations, so that the unmanned equipment can acquire acquired point cloud data, determine a corresponding undetermined detection range of the obstacles in the point cloud data, determine point cloud data in the undetermined detection range as target point cloud data, and then perform data enhancement on the target point cloud data based on standard point cloud data to obtain enhanced point cloud data, wherein the standard point cloud data is used for performing densification on the target point cloud data.
That is to say, the above-mentioned detection range to be decided may refer to a 3D range where the obstacle is located, which is preliminarily determined by the unmanned device with respect to the point cloud data, and the subsequent unmanned device may enhance the target point cloud data within the 3D range to obtain enhanced point cloud data, and determine a more accurate 3D range where the obstacle is located through the enhanced point cloud data.
The above undetermined detection range may be determined, and various data enhancement methods for the target point cloud data may be used, and the method for determining the undetermined detection range and enhancing the target point cloud data will be sequentially described below.
First, various ways to determine the range to be detected may be provided, for example, the unmanned device may determine a global feature corresponding to the point cloud data, determine a basic range of the obstacle in the point cloud data based on the global feature, determine a partial feature belonging to the basic range in the global feature, and further determine the range to be detected according to the partial feature. That is to say, the global feature may refer to a feature that represents the point cloud data as a whole, the range detection of the obstacle is performed through the global feature to determine a basic range in which the obstacle is approximately located in the point cloud data, and then, the undetermined detection range corresponding to the obstacle is determined in a refined manner through a partial feature that belongs to the basic range in the global feature, that is, a partial feature that belongs to the obstacle, and the point cloud data may relate to a plurality of obstacles.
When the undetermined detection range corresponding to the obstacle is determined through the partial feature corresponding to the obstacle, in order to more accurately determine the undetermined detection range, the candidate frames corresponding to the obstacle and the confidence degree corresponding to each candidate frame can be determined through the partial feature, each candidate frame corresponding to the obstacle can refer to a 3C frame which has different positions and sizes and represents the range of the obstacle in point cloud data, valuable candidate frames can be selected according to the confidence degree corresponding to each candidate frame, and the undetermined detection range can be determined according to the valuable candidate frames.
There may be various ways to pick out the candidate box. For example, the candidate box with the highest confidence may be selected as the target box. For another example, according to the coincidence rate and the confidence level between the candidate frames, the candidate frame with the high confidence level and the low coincidence rate may be selected, specifically, in the N-th round of screening, the candidate frame with the highest confidence level may be selected from the candidate frame set corresponding to the N-th round as the candidate frame screened by the N-th round, and the candidate frame with the highest confidence level in the candidate frame set corresponding to the N-th round exceeding the set coincidence rate is removed, so as to obtain the candidate frame set corresponding to the N + 1-th round until a preset screening condition is satisfied, and obtain the remaining candidate frames, where for any round of screening, if the candidate frame with the highest confidence level exceeding the set coincidence rate is not found from the candidate frame set corresponding to the wheel, the candidate frame set corresponding to the wheel is determined to satisfy the screening condition, and the candidate frame set obtained after the round of screening is used as the remaining candidate frame, and the target frame is determined according to the candidate frame screened by each round and the remaining candidate frame, where N is a positive integer.
It can be seen that the candidate frame set corresponding to each round is obtained in the previous round, and the candidate frame with the highest confidence coefficient having the higher coincidence rate than the set coincidence rate are taken out from the candidate frame set in the previous round, so that the candidate frame set required to be used in the next round is obtained. The set overlapping ratio mentioned here may be set in advance.
That is to say, the above-mentioned method is to perform the screening of the candidate frames for multiple times, and when screening the candidate frames each time, select the candidate frame with the highest confidence, and delete the candidate frame with the highest confidence and the candidate frame with the higher coincidence rate, so as to reach the purpose of selecting several candidate frames with higher confidence and lower coincidence rate, and thus take these candidate frames as the target frames. For example, assume that there are 4 candidate boxes for an obstacle: A. b, C and D, wherein the candidate frame set comprises [ A, B, C and D ], the confidence coefficients are gradually lower from A to D, when the candidate frame set is screened for the first time, the coincidence rate of D and A is higher than the set coincidence rate, and D is deleted and is not considered, so that the candidate frame set obtained in the round is [ B and C ], secondary screening based on the candidate frame set [ B and C ] is started, wherein the coincidence rate of C and B is also higher than the set coincidence rate, B can be left, therefore, A and B are selected target frames, and the undetermined detection range can refer to the range formed by A and B.
Of course, if the time for determining the undetermined detection range needs to be reduced, the unmanned device can also determine the undetermined detection range corresponding to the obstacle in the point cloud data directly through the global features of the point cloud data. The global feature may be determined in various ways, for example, the global feature of the point cloud data may be determined by a BEV algorithm, that is, the global feature may be referred to as a BEV2D feature.
Most of the above contents are used for explaining how to determine the detection range to be determined, and in the present specification, the main point for solving the problem in the prior art is to enhance the target point cloud data by normalizing the point cloud data, and perform range detection by the enhanced point cloud data to obtain a more accurate detection range than the detection range to be determined, that is, optimize the detection range. That is to say, the target point cloud data may have a certain problem of sparseness, deficiency, etc., and then the target point cloud data is filled by the standard point cloud data, so that a more accurate detection range after optimization can be obtained to a certain extent.
The normalized point cloud data may be a fixed density, and is uniformly distributed in and around the target point cloud data, as shown in fig. 2.
Fig. 2 is a schematic diagram of normalized point cloud data provided in this specification.
As can be seen from fig. 2, the standard point cloud data may be of a set density, that is, the standard point cloud data of the set density may be added to the target point cloud data to obtain enhanced point cloud data, in fig. 2, a black point represents an original point cloud point in the target point cloud data, a black frame represents a pending detection range, and a virtual point cloud composed of gray points may be of a fixed density.
It should be noted that, as can be seen from fig. 2, the standard point cloud data may exceed the undetermined detection range, and then when determining the enhanced point cloud data, the enhanced point cloud data may include the standard point cloud data and the target point cloud data within the undetermined detection range, and of course, the enhanced point cloud data may also include the standard point cloud data located outside the undetermined detection range in fig. 2, but the standard point cloud data added to the enhanced point cloud data needs to meet the requirement of distinguishing the edge point from the point cloud point outside the undetermined detection range.
Therefore, when the standard point cloud data is added to the target point cloud data, the standard point cloud data outside the undetermined detection range can be adjusted after the standard point cloud data is added to the target point cloud data, and/or the point cloud data in the undetermined detection range after the standard point cloud data is overlaid is adjusted, so that the enhanced point cloud data is obtained.
That is, the purpose of highlighting the edge of the target point cloud data can be achieved by adjusting the point cloud points outside the undetermined detection range and/or adjusting the point cloud points inside the undetermined detection range (for the point cloud inside the undetermined detection range, only the standard point cloud data can be adjusted, and both the standard point cloud data and the target point cloud data can be adjusted).
Based on this, the unmanned device can adjust the laser reflectivity of the standard point cloud data outside the undetermined detection range to be not more than the set reflectivity, wherein for any point cloud point, if the laser reflectivity of the point cloud point is higher, the point cloud point is more prominent in the point cloud, wherein the set reflectivity can be preset, for example, the laser reflectivity of the standard point cloud data outside the undetermined detection range can be adjusted to be 0, and the standard point cloud data added outside the undetermined detection range can be understood as the point cloud serving as the background of the target point cloud data.
In the undetermined detection range, the point cloud data in the undetermined detection range after the standard point cloud data is superimposed can be adjusted by using the point cloud data in the undetermined detection range, wherein the point cloud points closer to the center of the undetermined detection range have lower laser reflectivity, and the point cloud points closer to the edge of the undetermined detection range have higher laser reflectivity as an adjustment target.
That is to say, the laser reflectivity of the point cloud point closer to the boundary of the undetermined detection range is higher, and the laser reflectivity of the point cloud point farther from the boundary of the undetermined detection range is lower, so that the enhanced point cloud data can show the more obvious boundary of the obstacle corresponding to the target point cloud data, and the obstacle range detection is subsequently performed through the enhanced point cloud data, and the more accurate optimized detection range can be obtained.
S104: determining point cloud characteristics corresponding to the enhanced point cloud data, performing range detection on the obstacle according to the point cloud characteristics to obtain an optimized detection range, and performing obstacle detection on the obstacle based on the optimized detection range.
After the target point cloud data are subjected to data enhancement in the mode, the enhanced point cloud data are obtained, the unmanned equipment can determine point cloud characteristics corresponding to the enhanced point cloud data, range detection is carried out according to the point cloud characteristics, the undetermined detection range is optimized, and the optimized detection range is obtained.
That is to say, compared with the undetermined detection range, the optimized detection range is obtained by data enhancement of target point cloud data, and the undetermined detection range is obtained by point cloud data with certain defects (that is, point clouds are sparse, and the boundary may have defects such as missing).
It should be noted that, when determining the characteristics of the point cloud data, it is usually necessary to use coordinates of each point cloud point in the point cloud data, and the coordinates of the point cloud points are usually coordinates in a coordinate system centering on a collecting device (i.e., a radar installed on the unmanned device) for collecting the point cloud data, and thus, the coordinate values of each point cloud point in a portion of the point cloud corresponding to an obstacle farther away from the unmanned device may be larger, so that the coordinate values of each point cloud point of the portion of the point cloud may not represent a distance relationship between each point cloud point, which may also result in an inaccurate range of the determined obstacle, and when the unmanned device determines the range of the obstacle through such point cloud data, since the coordinate values are larger, the efficiency may also be lower, which will be described in the following manner to solve these problems.
Specifically, when determining the point cloud features of the enhanced point cloud data, an original coordinate corresponding to each point cloud point in the enhanced point cloud data may be determined, where the original coordinate is a coordinate in a coordinate system centered on a collecting device that collects the target point cloud data, and then the drone device may determine a reference point of the enhanced point cloud data (the reference point may be multiple, for example, the reference point may refer to a center point corresponding to the target point cloud data or may refer to any point cloud point in the target point cloud data), convert the original coordinate corresponding to the point cloud point into a coordinate in a coordinate system centered on the reference point, obtain a converted coordinate corresponding to the point cloud point, and determine the point cloud features corresponding to the enhanced point cloud data according to the converted coordinate corresponding to each point cloud point. (of course, since the reference point is the origin of coordinates, there is no need to convert the coordinates of the reference point, that is, for each cloud point except the reference point, the original coordinates corresponding to the cloud point are converted into converted coordinates)
That is to say, the coordinates of each point cloud point in the enhanced point cloud data can be converted into the coordinates in the local space of the enhanced point cloud data, that is, the coordinates of each point cloud point in the enhanced point cloud data are represented by a coordinate system at the position where the enhanced point cloud data is located, the numerical value of each coordinate is not large, and the positional relationship between the point cloud points can be more clearly represented by the relationship between the numerical values of the coordinates of each point cloud point.
After the optimized detection range is determined, obstacle detection can be performed on the obstacle based on the optimized detection range, and the unmanned equipment is controlled based on an obstacle detection result. That is, the unmanned aerial vehicle detects an obstacle with respect to the obstacle, for example, determines what obstacle the obstacle is, determines the position of the obstacle, and the like, based on the optimized detection range. After the obstacle detection result is obtained by performing obstacle detection, the unmanned device may be controlled based on the obstacle detection result, for example, after determining what kind of obstacle the obstacle is, the obstacle is specifically avoided. For another example, after the position of the obstacle is determined, the trajectory of the obstacle is predicted, and trajectory planning is performed based on the predicted trajectory of the obstacle.
The unmanned equipment mentioned above may refer to equipment capable of realizing automatic driving, such as unmanned vehicles, unmanned aerial vehicles, automatic distribution equipment, and the like. Therefore, the obstacle detection method for unmanned driving provided by the specification can be used for enhancing point cloud data by unmanned equipment and detecting obstacles by the enhanced point cloud data, and the unmanned equipment can be particularly applied to the field of delivery by the unmanned equipment, such as business scenes of delivery, logistics, takeaway and the like by using the unmanned equipment.
It should be noted that, in this specification, the networks for determining the undetermined detection range and determining the optimized detection range may be separate, and the network for determining the optimized detection range may be a lightweight network, and the number of network layers is small, and when determining the undetermined detection range, the point cloud data may be input into the first network, so as to obtain the undetermined detection range, and when determining the optimized detection range, the point cloud characteristics corresponding to the enhanced point cloud data may be determined through the second network, and according to the point cloud characteristics, the range detection is performed on the obstacle, so as to obtain the optimized detection range. Wherein the number of network layers of the second network is less than the number of network layers of the first network.
Wherein. The first network may include a first sub-network that determines the basic range and a second sub-network that determines the detection range to be determined, and after the basic range is determined by the first sub-network, the second sub-network may detect a part of the features belonging to the basic range to obtain the detection range to be determined.
The first network and the second network can be trained independently, and when the second network is trained, a training sample can be obtained, the training sample comprises obstacle point cloud data and a labeling range corresponding to the obstacle point cloud data, the obstacle point cloud data can be enhanced based on standard point cloud data and the labeling range, enhanced point cloud data corresponding to the obstacle point cloud data are obtained, the enhanced point cloud data are input into the second network to be trained, a prediction range is obtained, and the second network is trained by taking the deviation between the prediction range and the labeling range as a target.
That is, when the second network is trained, the obstacle point cloud data included in the training sample may be enhanced, and the range detection of the obstacle may be performed by determining the obstacle point cloud data based on the enhanced obstacle point cloud data through the second network, where the training target of the second network is to enable the prediction range obtained through the enhanced obstacle point cloud data to be accurate, so that when the second network is applied, the optimized prediction range may be obtained as accurately as possible through the second network.
According to the method, the standard point cloud data can be added into the target point cloud data corresponding to the obstacle based on the undetermined detection range, so that the enhanced point cloud data are densified, and part of missing positions are filled, so that the range of the obstacle can be more accurately determined through the enhanced point cloud data.
In addition, for the point cloud data after enhancement, the boundary of the point cloud representing the obstacle can be made obvious, the point cloud not belonging to the obstacle is weakened and used as the background of the point cloud setoff the obstacle, and therefore, the detection can be more accurate when the obstacle range detection is carried out through the point cloud data after enhancement.
Based on the same idea, the present specification further provides a corresponding obstacle detection device for unmanned driving, as shown in fig. 3.
Fig. 3 is a schematic diagram of an obstacle detection device for unmanned driving provided in this specification, which specifically includes:
the acquisition module 301 is configured to acquire point cloud data acquired by an unmanned device and determine a to-be-determined detection range of an obstacle in the point cloud data;
a determining module 302, configured to determine point cloud data in the pending detection range as target point cloud data;
an enhancement module 303, configured to perform data enhancement on the target point cloud data based on standard point cloud data to obtain enhanced point cloud data, where the standard point cloud data is used to perform densification on the target point cloud data;
and the optimization module 304 is configured to determine point cloud features corresponding to the enhanced point cloud data, perform range detection on the obstacle according to the point cloud features to obtain an optimized detection range, and perform obstacle detection on the obstacle based on the optimized detection range.
Optionally, the enhancing module 303 is specifically configured to add the standard point cloud data to the target point cloud data; and adjusting standard point cloud data outside the undetermined detection range, and/or adjusting point cloud data in the undetermined detection range after the standard point cloud data are superposed to obtain enhanced point cloud data.
Optionally, the enhancing module 303 is specifically configured to adjust a laser reflectivity of the standard point cloud data located outside the to-be-determined detection range to be not greater than a set reflectivity, where for any point cloud point, if the laser reflectivity of the point cloud point is higher, the point cloud point is more prominent in the point cloud.
Optionally, the enhancing module 303 is specifically configured to adjust the point cloud data in the undetermined detection range after the standard point cloud data is superimposed, with a target that laser reflectivity of a point cloud point in the point cloud data in the undetermined detection range that is closer to a center position of the undetermined detection range is lower and laser reflectivity of a point cloud point that is closer to an edge of the undetermined detection range is higher, where the point cloud data in the undetermined detection range after the standard point cloud data is superimposed is adjusted, and if the laser reflectivity of any point cloud point is higher, the point cloud point is more prominent in the point cloud.
Optionally, the optimization module 304 is specifically configured to determine an original coordinate corresponding to each cloud point in the enhanced point cloud data, where the original coordinate is a coordinate in a coordinate system centered on an acquisition device that acquires the target point cloud data; determining a reference point in the enhanced point cloud data, converting an original coordinate corresponding to each point cloud point of the enhanced point cloud data into a coordinate under a coordinate system taking the reference point as a center, and obtaining a converted coordinate corresponding to the point cloud point; and determining the point cloud characteristics corresponding to the enhanced point cloud data according to the converted coordinates corresponding to the cloud points of each point in the enhanced point cloud data.
Optionally, the obtaining module 301 is specifically configured to determine a global feature corresponding to the point cloud data, and determine a basic range of the obstacle in the point cloud data based on the global feature; determining partial features of the global features which belong to the basic range; and determining the undetermined detection range according to the partial characteristics.
Optionally, the obtaining module 301 is specifically configured to determine, according to the partial features, each candidate frame corresponding to the obstacle and a confidence level corresponding to each candidate frame; and selecting a target frame from the candidate frames according to the candidate frames and the confidence degrees corresponding to the candidate frames, and determining the to-be-determined detection range according to the target frame.
Optionally, the obtaining module 301 is specifically configured to, in the nth round of screening, screen a candidate frame with a highest confidence from the candidate frame set corresponding to the nth round as the candidate frame screened by the nth round, and remove a candidate frame with a highest confidence in the candidate frame set corresponding to the nth round, where the coincidence ratio with the candidate frame with the highest confidence exceeds a set coincidence ratio, to obtain a candidate frame set corresponding to the N +1 th round, until a preset screening condition is met, and obtain a remaining candidate frame, where, for any round of screening, if a candidate frame with a coincidence ratio with the candidate frame screened by the round that exceeds the set coincidence ratio is not found from the candidate frame set corresponding to the round, it is determined that the screening condition is met, and the candidate frame set obtained after the round of screening is used as the remaining candidate frame, where N is a positive integer; and determining the target frame according to the candidate frame screened out in each round and the residual candidate frame.
Optionally, the obtaining module 301 is specifically configured to input the point cloud data into a first network, and determine an undetermined detection range of an obstacle corresponding to the point cloud data; the optimization module 304 is specifically configured to determine, through a second network, a point cloud feature corresponding to the enhanced point cloud data, and perform range detection on the obstacle according to the point cloud feature to obtain an optimized detection range, where the number of network layers of the second network is smaller than the number of network layers of the first network.
Optionally, the apparatus further comprises:
a training module 305, configured to obtain a training sample, where the training sample includes obstacle point cloud data and a labeling range corresponding to the obstacle point cloud data; based on the standard point cloud data and the labeling range, enhancing the obstacle point cloud data to obtain enhanced point cloud data corresponding to the obstacle point cloud data; inputting the enhanced point cloud data into a second network to be trained to obtain a prediction range; and training the second network by taking the deviation between the prediction range and the labeling range as a target to be minimized.
The present specification also provides a computer-readable storage medium storing a computer program operable to execute the obstacle detection method for unmanned driving provided in fig. 1 described above.
This description also provides a schematic block diagram of the drone shown in figure 4. As shown in fig. 4, the drone includes, at the hardware level, a processor, an internal bus, a network interface, a memory, and a non-volatile memory, although it may also include hardware required for other services. The processor reads a corresponding computer program from the non-volatile memory into the memory and then runs the computer program to implement the obstacle detection method for unmanned driving described above with reference to fig. 1. Of course, besides the software implementation, this specification does not exclude other implementations, such as logic devices or combination of software and hardware, and so on, that is, the execution subject of the following processing flow is not limited to each logic unit, and may be hardware or logic devices.
In the 90 s of the 20 th century, improvements in a technology could clearly distinguish between improvements in hardware (e.g., improvements in circuit structures such as diodes, transistors, switches, etc.) and improvements in software (improvements in process flow). However, as technology advances, many of today's process flow improvements have been seen as direct improvements in hardware circuit architecture. Designers almost always obtain the corresponding hardware circuit structure by programming an improved method flow into the hardware circuit. Thus, it cannot be said that an improvement in the process flow cannot be realized by hardware physical blocks. For example, a Programmable Logic Device (PLD), such as a Field Programmable Gate Array (FPGA), is an integrated circuit whose Logic functions are determined by programming the Device by a user. A digital system is "integrated" on a PLD by the designer's own programming without requiring the chip manufacturer to design and fabricate application-specific integrated circuit chips. Furthermore, nowadays, instead of manually manufacturing an Integrated Circuit chip, such Programming is often implemented by "logic compiler" software, which is similar to a software compiler used in program development, but the original code before compiling is also written in a specific Programming Language, which is called Hardware Description Language (HDL), and the HDL is not only one kind but many kinds, such as abll (Advanced boot Expression Language), AHDL (alternate hard Description Language), traffic, CUPL (computer universal Programming Language), HDCal (Java hard Description Language), lava, lola, HDL, PALASM, software, rhydl (Hardware Description Language), and vhul-Language (vhyg-Language), which is currently used in the field. It will also be apparent to those skilled in the art that hardware circuitry that implements the logical method flows can be readily obtained by merely slightly programming the method flows into an integrated circuit using the hardware description languages described above.
The controller may be implemented in any suitable manner, for example, the controller may take the form of, for example, a microprocessor or processor and a computer readable medium that stores computer readable program code (e.g., software or firmware) executable by the (micro) processor, logic gates, switches, an Application Specific Integrated Circuit (ASIC), a programmable logic controller, and embedded microcontrollers, examples of which include, but are not limited to, the following microcontrollers: ARC 625D, atmel AT91SAM, microchip PIC18F26K20, and Silicone Labs C8051F320, the memory controller may also be implemented as part of the control logic for the memory. Those skilled in the art will also appreciate that, in addition to implementing the controller as pure computer readable program code, the same functionality can be implemented by logically programming method steps such that the controller is in the form of logic gates, switches, application specific integrated circuits, programmable logic controllers, embedded microcontrollers and the like. Such a controller may thus be considered a hardware component, and the means included therein for performing the various functions may also be considered as a structure within the hardware component. Or even means for performing the functions may be regarded as being both a software module for performing the method and a structure within a hardware component.
The systems, apparatuses, modules or units described in the above embodiments may be specifically implemented by a computer chip or an entity, or implemented by a product with certain functions. One typical implementation device is a computer. In particular, the computer may be, for example, a personal computer, a laptop computer, a cellular telephone, a camera phone, a smartphone, a personal digital assistant, a media player, a navigation device, an email device, a game console, a tablet computer, a wearable device, or a combination of any of these devices.
For convenience of description, the above devices are described as being divided into various units by function, and are described separately. Of course, the functions of the various elements may be implemented in the same one or more software and/or hardware implementations of the present description.
As will be appreciated by one skilled in the art, embodiments of the present invention may be provided as a method, system, or computer program product. Accordingly, the present invention may take the form of an entirely hardware embodiment, an entirely software embodiment or an embodiment combining software and hardware aspects. Furthermore, the present invention may take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, and the like) having computer-usable program code embodied therein.
The present invention is described with reference to flowchart illustrations and/or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the invention. It will be understood that each flow and/or block of the flowchart illustrations and/or block diagrams, and combinations of flows and/or blocks in the flowchart illustrations and/or block diagrams, can be implemented by computer program instructions. These computer program instructions may be provided to a processor of a general purpose computer, special purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, create means for implementing the functions specified in the flowchart flow or flows and/or block diagram block or blocks.
These computer program instructions may also be stored in a computer-readable memory that can direct a computer or other programmable data processing apparatus to function in a particular manner, such that the instructions stored in the computer-readable memory produce an article of manufacture including instruction means which implement the function specified in the flowchart flow or flows and/or block diagram block or blocks.
These computer program instructions may also be loaded onto a computer or other programmable data processing apparatus to cause a series of operational steps to be performed on the computer or other programmable apparatus to produce a computer implemented process such that the instructions which execute on the computer or other programmable apparatus provide steps for implementing the functions specified in the flowchart flow or flows and/or block diagram block or blocks.
In a typical configuration, a computing device includes one or more processors (CPUs), input/output interfaces, network interfaces, and memory.
The memory may include forms of volatile memory in a computer readable medium, random Access Memory (RAM) and/or non-volatile memory, such as Read Only Memory (ROM) or flash memory (flash RAM). Memory is an example of a computer-readable medium.
Computer-readable media, including both non-transitory and non-transitory, removable and non-removable media, may implement information storage by any method or technology. The information may be computer readable instructions, data structures, modules of a program, or other data. Examples of computer storage media include, but are not limited to, phase change memory (PRAM), static Random Access Memory (SRAM), dynamic Random Access Memory (DRAM), other types of Random Access Memory (RAM), read Only Memory (ROM), electrically Erasable Programmable Read Only Memory (EEPROM), flash memory or other memory technology, compact disc read only memory (CD-ROM), digital Versatile Discs (DVD) or other optical storage, magnetic cassettes, magnetic tape magnetic disk storage or other magnetic storage devices, or any other non-transmission medium that can be used to store information that can be accessed by a computing device. As defined herein, a computer readable medium does not include a transitory computer readable medium such as a modulated data signal and a carrier wave.
It should also be noted that the terms "comprises," "comprising," or any other variation thereof, are intended to cover a non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements does not include only those elements but may include other elements not expressly listed or inherent to such process, method, article, or apparatus. Without further limitation, an element defined by the phrases "comprising one of 8230; \8230;" 8230; "does not exclude the presence of additional like elements in a process, method, article, or apparatus that comprises that element.
As will be appreciated by one skilled in the art, embodiments of the present description may be provided as a method, system, or computer program product. Accordingly, the description may take the form of an entirely hardware embodiment, an entirely software embodiment or an embodiment combining software and hardware aspects. Furthermore, the description may take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, and so forth) having computer-usable program code embodied therein.
This description may be described in the general context of computer-executable instructions, such as program modules, being executed by a computer. Generally, program modules include routines, programs, objects, components, data structures, etc. that perform particular tasks or implement particular abstract data types. The specification may also be practiced in distributed computing environments where tasks are performed by remote processing devices that are linked through a communications network. In a distributed computing environment, program modules may be located in both local and remote computer storage media including memory storage devices.
The embodiments in the present specification are described in a progressive manner, and the same and similar parts among the embodiments are referred to each other, and each embodiment focuses on the differences from the other embodiments. In particular, for the system embodiment, since it is substantially similar to the method embodiment, the description is simple, and for the relevant points, reference may be made to the partial description of the method embodiment.
The above description is only an example of the present specification, and is not intended to limit the present specification. Various modifications and alterations to this description will become apparent to those skilled in the art. Any modification, equivalent replacement, improvement or the like made within the spirit and principle of the present specification should be included in the scope of the claims of the present specification.
Claims (12)
1. An obstacle detection method for unmanned driving, characterized by comprising:
acquiring point cloud data acquired by unmanned equipment, and determining a corresponding undetermined detection range of an obstacle in the point cloud data;
determining point cloud data in the undetermined detection range as target point cloud data;
based on standard point cloud data and the undetermined detection range, performing data enhancement on the target point cloud data to obtain enhanced point cloud data, wherein the standard point cloud data is used for performing densification and missing filling on the target point cloud data;
determining point cloud features corresponding to the enhanced point cloud data, performing range detection on the obstacle according to the point cloud features to obtain an optimized detection range, and performing obstacle detection on the obstacle based on the optimized detection range;
the method comprises the steps of acquiring point cloud data acquired by unmanned equipment, determining a corresponding undetermined detection range of an obstacle in the point cloud data, and specifically comprising the following steps:
inputting the point cloud data into a first network, and determining a corresponding undetermined detection range of the obstacle in the point cloud data;
determining point cloud characteristics corresponding to the enhanced point cloud data, and performing range detection on the obstacle according to the point cloud characteristics to obtain an optimized detection range, wherein the method specifically comprises the following steps:
and determining point cloud characteristics corresponding to the enhanced point cloud data through a second network, and performing range detection on the barrier according to the point cloud characteristics to obtain an optimized detection range, wherein the number of network layers of the second network is smaller than that of the first network.
2. The method of claim 1, wherein the data enhancement of the target point cloud data is performed based on the normative point cloud data and the undetermined detection range to obtain enhanced point cloud data, and specifically comprises:
adding the standard point cloud data to the target point cloud data;
and adjusting standard point cloud data outside the undetermined detection range, and/or adjusting point cloud data in the undetermined detection range after the standard point cloud data are superposed to obtain enhanced point cloud data.
3. The method of claim 2, wherein adjusting canonical point cloud data that is outside the range to be detected specifically comprises:
and adjusting the laser reflectivity of the standard point cloud data outside the undetermined detection range to be not more than the set reflectivity, wherein for any point cloud point, if the laser reflectivity of the point cloud point is higher, the point cloud point is more prominent in the point cloud.
4. The method of claim 2, wherein adjusting the point cloud data within the range to be detected after superimposing the canonical point cloud data comprises:
adjusting the point cloud data in the undetermined detection range after the standard point cloud data is superposed, wherein the point cloud data is the more prominent the point cloud point in the point cloud if the laser reflectivity of the point cloud point is higher for any point cloud point.
5. The method of claim 1, wherein determining the point cloud features corresponding to the enhanced point cloud data specifically comprises:
determining an original coordinate corresponding to each point cloud point in the enhanced point cloud data, wherein the original coordinate is a coordinate under a coordinate system taking acquisition equipment for acquiring the target point cloud data as a center;
determining a reference point in the enhanced point cloud data, converting an original coordinate corresponding to each point cloud point of the enhanced point cloud data into a coordinate under a coordinate system taking the reference point as a center, and obtaining a converted coordinate corresponding to the point cloud point;
and determining the point cloud characteristics corresponding to the enhanced point cloud data according to the converted coordinates corresponding to the cloud points of each point in the enhanced point cloud data.
6. The method of claim 1, wherein determining a corresponding pending detection range of an obstacle in the point cloud data comprises:
determining global features corresponding to the point cloud data, and determining a basic range of the obstacle in the point cloud data based on the global features;
determining partial features of the global features which belong to the basic range;
and determining the undetermined detection range according to the partial characteristics.
7. The method of claim 6, wherein determining the range to be detected based on the partial features comprises:
determining each candidate frame corresponding to the obstacle and the confidence corresponding to each candidate frame according to the partial features;
and selecting a target frame from the candidate frames according to the candidate frames and the confidence degrees corresponding to the candidate frames, and determining the range to be detected according to the target frame.
8. The method of claim 7, wherein selecting a target frame from the candidate frames according to the candidate frames and the confidence levels corresponding to the candidate frames comprises:
in the N round of screening, screening a candidate frame with the highest confidence coefficient from a candidate frame set corresponding to the N round as a candidate frame screened by the N round, and removing a candidate frame, of which the coincidence rate with the candidate frame with the highest confidence coefficient exceeds a set coincidence rate, from the candidate frame set corresponding to the N round to obtain a candidate frame set corresponding to the (N + 1) th round until a preset screening condition is met, and obtaining a remaining candidate frame, wherein for any round of screening, if a candidate frame, of which the coincidence rate with the candidate frame screened by the round exceeds the set coincidence rate, is not found from the candidate frame set corresponding to the round, the candidate frame set meeting the screening condition is determined, and the candidate frame set obtained after the round of screening is used as the remaining candidate frame, wherein N is a positive integer;
and determining the target frame according to the candidate frames screened out in each round and the residual candidate frames.
9. The method of claim 1, wherein training the second network specifically comprises:
acquiring a training sample, wherein the training sample comprises obstacle point cloud data and a labeling range corresponding to the obstacle point cloud data;
enhancing the obstacle point cloud data based on the standard point cloud data and the labeling range to obtain enhanced point cloud data corresponding to the obstacle point cloud data;
inputting the enhanced point cloud data into a second network to be trained to obtain a prediction range;
and training the second network by taking the deviation between the prediction range and the labeling range as a target to be minimized.
10. An obstacle detection device for unmanned driving, comprising:
the system comprises an acquisition module, a detection module and a detection module, wherein the acquisition module is used for acquiring point cloud data acquired by unmanned equipment and determining a corresponding undetermined detection range of an obstacle in the point cloud data;
the determining module is used for determining point cloud data in the undetermined detection range as target point cloud data;
the enhancement module is used for carrying out data enhancement on the target point cloud data based on standard point cloud data to obtain enhanced point cloud data, and the standard point cloud data is used for carrying out densification on the target point cloud data;
the optimization module is used for determining point cloud characteristics corresponding to the enhanced point cloud data, performing range detection on the obstacle according to the point cloud characteristics to obtain an optimized detection range, and performing obstacle detection on the obstacle based on the optimized detection range;
the acquisition module is used for inputting the point cloud data into a first network and determining a corresponding undetermined detection range of the obstacle in the point cloud data;
and the optimization module is used for determining point cloud characteristics corresponding to the enhanced point cloud data through a second network, and performing range detection on the barrier according to the point cloud characteristics to obtain an optimized detection range, wherein the number of network layers of the second network is smaller than that of the first network.
11. A computer-readable storage medium, characterized in that the storage medium stores a computer program which, when executed by a processor, implements the method of any of the preceding claims 1 to 9.
12. An unmanned aerial vehicle comprising a memory, a processor and a computer program stored on the memory and executable on the processor, wherein the processor when executing the program implements the method of any of claims 1 to 9.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202111104226.2A CN113887351B (en) | 2021-09-22 | 2021-09-22 | Obstacle detection method and obstacle detection device for unmanned driving |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202111104226.2A CN113887351B (en) | 2021-09-22 | 2021-09-22 | Obstacle detection method and obstacle detection device for unmanned driving |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN113887351A CN113887351A (en) | 2022-01-04 |
| CN113887351B true CN113887351B (en) | 2023-03-07 |
Family
ID=79010111
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202111104226.2A Active CN113887351B (en) | 2021-09-22 | 2021-09-22 | Obstacle detection method and obstacle detection device for unmanned driving |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN113887351B (en) |
Families Citing this family (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN116740114B (en) * | 2023-08-11 | 2023-11-21 | 之江实验室 | Object boundary fitting method and device based on convex hull detection |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108152831A (en) * | 2017-12-06 | 2018-06-12 | 中国农业大学 | A kind of laser radar obstacle recognition method and system |
| CN110096059A (en) * | 2019-04-25 | 2019-08-06 | 杭州飞步科技有限公司 | Automatic Pilot method, apparatus, equipment and storage medium |
| CN112330815A (en) * | 2020-11-26 | 2021-02-05 | 北京百度网讯科技有限公司 | Three-dimensional point cloud data processing method, device and equipment based on obstacle fusion |
| CN112712009A (en) * | 2020-12-28 | 2021-04-27 | 北京三快在线科技有限公司 | Method and device for detecting obstacle |
Family Cites Families (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109101861A (en) * | 2017-06-20 | 2018-12-28 | 百度在线网络技术（北京）有限公司 | Obstacle identity recognition methods, device, equipment and storage medium |
| CN112677993A (en) * | 2021-01-05 | 2021-04-20 | 北京三快在线科技有限公司 | Model training method and device |
-
2021
- 2021-09-22 CN CN202111104226.2A patent/CN113887351B/en active Active
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108152831A (en) * | 2017-12-06 | 2018-06-12 | 中国农业大学 | A kind of laser radar obstacle recognition method and system |
| CN110096059A (en) * | 2019-04-25 | 2019-08-06 | 杭州飞步科技有限公司 | Automatic Pilot method, apparatus, equipment and storage medium |
| CN112330815A (en) * | 2020-11-26 | 2021-02-05 | 北京百度网讯科技有限公司 | Three-dimensional point cloud data processing method, device and equipment based on obstacle fusion |
| CN112712009A (en) * | 2020-12-28 | 2021-04-27 | 北京三快在线科技有限公司 | Method and device for detecting obstacle |
Also Published As
| Publication number | Publication date |
|---|---|
| CN113887351A (en) | 2022-01-04 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN111311709B (en) | Method and device for generating high-precision map | |
| CN111639682A (en) | Ground segmentation method and device based on point cloud data | |
| CN111797711A (en) | Model training method and device | |
| CN115600157B (en) | Data processing method and device, storage medium and electronic equipment | |
| CN113887351B (en) | Obstacle detection method and obstacle detection device for unmanned driving | |
| CN112365513A (en) | Model training method and device | |
| CN112990099B (en) | Method and device for detecting lane line | |
| CN116721399B (en) | Point cloud target detection method and device for quantitative perception training | |
| CN111476729B (en) | Target identification method and device | |
| CN112818968A (en) | Target object classification method and device | |
| CN111426299B (en) | Method and device for ranging based on depth of field of target object | |
| CN115880685A (en) | Three-dimensional target detection method and system based on votenet model | |
| CN115984154A (en) | Image fusion method and device, storage medium and electronic equipment | |
| CN112734851B (en) | Pose determination method and device | |
| CN117095371A (en) | Target detection method and detection device | |
| CN114863206A (en) | Model training method, target detection method and device | |
| CN112712561A (en) | Picture construction method and device, storage medium and electronic equipment | |
| CN112036462B (en) | Model training and target detection method and device | |
| CN112329547A (en) | Data processing method and device | |
| CN116740197B (en) | External parameter calibration method and device, storage medium and electronic equipment | |
| CN117746193B (en) | Label optimization method and device, storage medium and electronic equipment | |
| CN116363390B (en) | Infrared dim target detection method and device, storage medium and electronic equipment | |
| CN113205144B (en) | Model training method and device | |
| CN113640823B (en) | Method and device for map drawing based on laser reflectivity base map | |
| CN116152274B (en) | Image segmentation method, device, equipment and storage medium |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |