CN112669458A - Method, device and program carrier for ground filtering based on laser point cloud - Google Patents
Method, device and program carrier for ground filtering based on laser point cloud Download PDFInfo
- Publication number
- CN112669458A CN112669458A CN202011559390.8A CN202011559390A CN112669458A CN 112669458 A CN112669458 A CN 112669458A CN 202011559390 A CN202011559390 A CN 202011559390A CN 112669458 A CN112669458 A CN 112669458A
- Authority
- CN
- China
- Prior art keywords
- point cloud
- ground
- laser point
- model
- cloud data
- 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.)
- Pending
Links
Images
Abstract
The present invention relates to the field of image processing. The invention provides a ground filtering method based on laser point cloud, which comprises the following steps: s1: acquiring laser point cloud data; s2: estimating a ground model from the laser point cloud data based on model fitting; and S3: and verifying the estimated ground model based on the feature comparison. The invention also provides equipment for ground filtering based on the laser point cloud, a laser radar system and a machine readable program carrier. The invention aims to combine the ground model estimation based on model fitting with the verification process based on feature comparison so as to improve the accuracy of the whole model estimation, particularly achieve the optimized fitting effect on the uneven ground and obviously improve the calculation efficiency.
Description
Technical Field
The invention relates to a method for ground filtering based on laser point cloud, equipment for ground filtering based on laser point cloud, a laser radar system and a machine readable program carrier.
Background
The laser radar plays an excellent role in the field of automatic driving, and multiple tasks such as vehicle positioning assistance, high-precision map building, obstacle detection, target tracking and the like can be realized by processing the obtained laser point cloud. To better accomplish these tasks, it is often necessary to perform ground filtering or segmentation based on the laser point cloud.
There are two main ground filtering methods in the prior art, one is a ground filtering method based on model fitting, and the method usually performs ground plane estimation on the whole laser point cloud, so that accurate fitting of local non-flat or inclined ground cannot be well realized. The other method is a feature-based ground fitting method, which uses a local convexity standard to classify laser point cloud data in a non-flat environment, but the method has high calculation cost and is often specific to a special ground form, so that the method cannot be suitable for large-area ground segmentation and has poor robustness.
Disclosure of Invention
The invention aims to provide a method for performing ground filtering based on laser point cloud, equipment for performing ground filtering based on laser point cloud, a laser radar system and a machine readable program carrier, so as to solve at least part of problems in the prior art.
According to a first aspect of the present invention, a method for ground filtering based on a laser point cloud is provided, the method comprising the following steps:
s1: acquiring laser point cloud data;
s2: estimating a ground model from the laser point cloud data based on model fitting; and
s3: and verifying the estimated ground model based on the feature comparison.
The invention comprises the following technical advantages: according to the method, a ground segmentation technology based on model fitting and an error correction technology based on the characteristic vector are combined, so that the inaccuracy of ground model estimation is overcome, and meanwhile, the calculation cost is reduced. Therefore, good effect of laser point cloud classification can be achieved at flat terrain, overfitting or underfitting errors are reduced at inclined road surfaces or depressions, and loss of terrain points or mistaken classification of points on other environmental objects into a ground model is favorably avoided.
Optionally, before step S2, additionally performing: the complete laser point cloud is divided, in particular non-uniformly divided, into a plurality of grid regions according to the region positions and sizes in the laser point cloud. The whole laser point cloud is gridded, and the ground model is estimated independently for each grid area, so that the accuracy of the estimation of the whole model can be improved, the optimal fitting effect can be particularly achieved on the uneven ground, and meanwhile, the calculation efficiency can be obviously improved.
Optionally, the method further comprises: the size of the grid area is determined in dependence on the distance to the lidar sensor, preferably the size of the grid area is increased with increasing distance to the lidar sensor. The following technical concept is used in particular: the laser radar sensor emits and receives laser pulse signals in a radioactive mode, points in the laser point cloud which are far away from the laser radar sensor are often more sparse, and points which are close to the laser radar sensor are more dense. In order to dynamically adapt the size of the grid area to the distribution and density of the points in the laser point cloud, the size of the grid area can be adjusted according to the distance relative to the laser radar sensor, so that a foundation is laid for the subsequent rational estimation of the ground model. Furthermore, it is also conceivable to adjust the size of the grid area based on the configuration and parameters of the lidar sensor used.
Optionally, the step S2 includes: defining seed nodes in the laser point cloud data; and judging whether the seed node and the neighborhood node have continuity or not, and classifying the seed node and the neighborhood node into the same plane model under the condition of continuity. In particular, position coordinates can be assigned to the individual nodes in the laser point cloud, and the position coordinates of the individual nodes can therefore be assigned to the respective ground model according to the relationship between coordinates and points.
Optionally, the step S2 includes: selecting a subset of laser point cloud data; establishing an initial ground model based on points in the subset of the laser point cloud data; and comparing the distance of further points in the laser point cloud data with respect to the initial ground model to a predefined distance threshold, wherein points whose distance with respect to the initial ground model is below the distance threshold are classified as ground points and points whose distance with respect to the initial ground model exceeds the distance threshold are classified as non-ground points.
Optionally, the step S2 includes: random sampling fitting is carried out on the laser point cloud data by means of a RANSAC algorithm, and an optimal ground model is estimated through iterative calculation.
Optionally, the step S3 includes:
calculating a feature vector for each point in the laser point cloud data;
calculating a reference vector for the ground model estimated in step S2;
comparing the feature vector of each point with the reference vector; and
false positive errors and false negative errors in the ground model estimate are rejected based on the results of the comparison.
In the method, abnormal points in the preliminary estimation of the ground model can be effectively removed through calculation and comparison based on the feature vectors, and the defect of insufficient filtering effect under the condition of complex terrain is favorably overcome.
Optionally, the feature vector comprises a normal vector, a directional gradient vector and/or a height gradient vector.
According to a second aspect of the invention, there is provided an apparatus for ground filtering based on a laser point cloud, the apparatus being adapted to perform the method according to the first aspect of the invention, the apparatus comprising:
an acquisition module configured to be able to acquire laser point cloud data;
an estimation module configured to estimate a ground model from the laser point cloud data based on a model fit; and
a verification module configured to enable verification of the estimated ground model based on the feature comparison.
According to a third aspect of the present invention, there is provided a lidar system comprising:
a lidar sensor configured to be capable of transmitting and receiving radar signals;
a signal processing device configured to be capable of processing a lidar signal of the lidar sensor and for outputting laser point cloud data; and
the apparatus according to the second aspect of the invention.
According to a fourth aspect of the present invention, there is provided a machine readable program carrier having stored thereon a computer program for performing the method according to the first aspect of the present invention when the computer program runs on a computer.
Drawings
The principles, features and advantages of the present invention may be better understood by describing the invention in more detail below with reference to the accompanying drawings. The drawings comprise:
FIG. 1 shows a flow diagram of a method for ground filtering based on a laser point cloud in accordance with an exemplary embodiment of the present invention;
fig. 2 shows a schematic representation of the gridding of a laser point cloud by means of the method according to the invention;
FIGS. 3a and 3b show schematic diagrams of false positive errors and false negative errors, respectively, in rejection model estimation in connection with a method according to the invention;
FIG. 4 shows a block diagram of an apparatus for ground filtering based on a laser point cloud according to an exemplary embodiment of the present invention; and
fig. 5 shows a block diagram of a lidar system according to an exemplary embodiment of the invention.
Detailed Description
In order to make the technical problems, technical solutions and advantageous effects of the present invention more apparent, the present invention will be described in further detail with reference to the accompanying drawings and exemplary embodiments. It should be understood that the specific embodiments described herein are merely illustrative of the invention and do not limit the scope of the invention.
Fig. 1 shows a flow chart of a method for ground filtering based on a laser point cloud according to an exemplary embodiment of the present invention.
In step S1, laser point cloud data is acquired. In this case, for example, the surroundings can be scanned by means of a lidar sensor to obtain 3D point cloud data, or the laser point cloud data can be received from a lidar system by means of a communication interface. After simple preprocessing, the laser point cloud data can be assigned position information, in particular can be represented as three-dimensional coordinates in a cartesian coordinate system.
In step S2: a ground model is estimated from the laser point cloud data based on model fitting.
As an example, after the laser point cloud data has been obtained, the complete laser point cloud can be divided into a plurality of mesh regions by the mesh pavement model according to the region positions and sizes in the laser point cloud, and the ground model estimation can be performed separately for each mesh region. Thereby, the temporal or spatial relationship between the individual points in the laser point cloud is determined.
By way of example, a seed node may be defined, and then, according to connectivity between the seed node and its neighborhood node, the topological connection relationship between the two nodes is deleted or retained according to a spatial region growing condition, so that the seed node can be expanded outwards to add the neighborhood node to its own plane model.
As an example, random Sample fitting may be performed on the laser point cloud data by means of a ransac (random Sample consensus) algorithm, and an optimal ground model may be estimated by iterative calculations. The input to the RANSAC algorithm is a set of raw observations of the laser point cloud, which typically contain large noise and null points. RANSAC realizes random sampling fitting by repeatedly selecting random subsets in laser point cloud data, and finally obtains a fitting plane close to an ideal model.
As an example, the best ground model is iteratively estimated by means of the RANSAC algorithm as follows:
-randomly selecting a subset in the originally acquired laser point cloud dataset, assuming points in the subset as belonging to the ground model, while designating points in the subset as local points;
-building an initial mathematical model of the ground based on the sampled local interior points, i.e. all unknown parameters of the initial mathematical model can be calculated from the assumed local interior points;
-determining whether further points in the laser point cloud dataset fulfill the initial mathematical model according to a preset threshold, if so marking the further points as intra-local points and recording the current number of intra-local points;
-repeating the iteration a number of times, updating the mathematical model of the ground based on the number of local points and the performance of the re-established model;
exit the iterative loop when the iteration exit condition is satisfied, thus resulting in the best solution (best-fit ground model) throughout the iteration.
As an example, when determining local points in the iterative process described above, a distance between the selected point and the estimated ground mathematical model may be calculated and compared to a predefined distance threshold, wherein points below the distance threshold are classified as ground points and points above the distance threshold are classified as non-ground points.
In step S3, the estimated ground model is verified based on the feature comparison.
In this case, the verification process can be carried out, for example, as follows: first, feature vectors, which may be, for example, normal vectors and/or directional gradient vectors, are calculated for each point in the laser point cloud data. Reference vectors may then be calculated for the ground model estimated in step S2, where they may be either considered as a whole or may be calculated separately for each mesh region, for example. The feature vectors of the individual points are compared with the reference vector and false positive and false negative errors in the ground model estimation are rejected based on the result of the comparison.
As an example, normal vectors are calculated for those points that have been classified as ground points (i.e., points whose distance from the ground mathematical model is within a distance threshold), while normal vectors are calculated for the ground model on which the point is based. The normal vectors calculated for the ground points are compared with the normal vectors calculated for the ground model, in which case inter alia the angle between them can be calculated, which means that these ground points belong to "false positive errors" in the ground model estimation process and should be rejected from the ground points if the angle between them is greater than an angle threshold (for example up to 25 °). Conversely, if the included angle is less than the angle threshold, it indicates that the ground points are correctly divided and should continue to be retained.
As an example, the non-ground points of its neighborhood (i.e., those points whose distance from the ground mathematical model is outside of a distance threshold) may be traversed starting from any ground point on the ground model while calculating the height gradient between that ground point and the adjacent non-ground points, where, for example, the coordinates of that ground point and non-ground point, respectively, in a cartesian coordinate system are first determined. The height gradient may then be determined as the height difference between the two points divided by the horizontal distance between the two points. Next, the resulting height gradient is compared to a gradient threshold, and if the height gradient is less than the gradient threshold, it indicates that the selected non-ground points belong to a "false negative error" in the ground model estimation, and therefore should be removed from the non-ground points and added to the ground points. Conversely, if the resulting height gradient is greater than the gradient threshold, it indicates that the non-ground point is correctly classified and should continue to be retained.
Fig. 2 shows a schematic illustration of the gridding of a laser point cloud by means of a method according to an embodiment of the invention.
A diagram of a projection of an originally acquired laser point cloud 201, comprising a plurality of points 203, onto a two-dimensional plane is shown at the top of fig. 2. The result 202 after performing the meshing on the laser point cloud is shown in the lower part of fig. 2. Here, the complete laser point cloud 201 may be divided non-uniformly into a plurality of grid areas 204, for example, according to the area position and size in the laser point cloud. As shown in fig. 2, the lower mesh region 204 is denser than the upper mesh region 204 because the size of each mesh region is determined according to the distance from the lidar sensor (arranged below the image in a manner not shown) in the present embodiment, in which the size of the mesh region is increased as the distance from the lidar sensor increases.
Before the laser point cloud is subjected to gridding processing, a Neighborhood map (neighborwood Graph) of the laser point cloud can be constructed in advance, so that the determination of the topological relation among all points in the point cloud is realized. Common methods of constructing neighborhood maps of laser point clouds include space planning partitioning, octree partitioning, and K-D tree partitioning. As an example, a neighborhood graph may be generated based on the scanning sequence of the lidar and the physical setup of the scanner, where a seed node is defined in the laser point cloud data, and then the seed node is assigned neighborhood nodes in a certain order and direction. Illustratively, neighborhood nodes positioned in four directions, namely, up, down, left and right, of the seed node respectively represent: during the scanning process of the laser radar, the node to be scanned next (left), the node scanned just before (right), the node scanned by the upper laser diode with the closest emission angle (upper), and the node scanned by the lower laser diode with the closest emission angle (lower).
Fig. 3a and 3b show schematic diagrams of false positive errors and false negative errors, respectively, in a culling model estimation in combination with a method according to an embodiment of the invention.
A schematic diagram of the estimated local ground model for each mesh is shown at the top of fig. 3 a. Ground points p1 and p2 are illustratively chosen within the grid 304 and their normal vectors 301, 302 are computed, respectively. Next, normal vectors 303 of the local ground model corresponding to the mesh 304 are determined (i.e., normal vectors of the fitted surface are determined).
To screen out false positive errors, it can be individually determined whether the orientation of the ground point is within a threshold range of the reference direction. In the present embodiment, it is determined whether the normal vector to the ground point p1, p2 makes an angle with the surface normal vector 303 of the associated ground model that is less than an angular threshold (e.g., 25 °).
A schematic diagram of the results of a comparison of the normal vectors 301, 302 of the ground points p1, p2 with the normal vector 303 of the local ground model is shown in the lower part of fig. 3 a. Here, it can be seen that an angle between the normal vector 301 of the ground point p1 and the normal vector 303 of the local ground model is θ 1, and an angle between the normal vector 302 of the ground point p2 and the normal vector 303 of the local ground model is θ 2, where θ 1 is greater than an angle threshold θ 0, and θ 2 is smaller than the angle threshold θ 0. This means that point p1 belongs to a false positive error generated in the ground model estimation (e.g. p1 may belong to a point on some object close to the ground) and should be removed from the estimated ground model, while point p2 belongs to a correctly classified point and remains as a ground point for the rest.
The upper part of fig. 3b also shows a schematic view of the estimated local ground model for each mesh. A ground point p3 is illustratively selected within the grid 304, while neighborhood nodes p4 and p5 that have been classified as non-ground points are selected from the ground point p 3.
To screen out false negative errors, it may be determined whether the height gradient component between the ground point p3 and the non-ground points p4, p5 is below a threshold. The height gradient or slope between the ground point p3 and the non-ground points p4, p5 is shown below fig. 3 b. Here, the height gradient (i.e., slope) is determined as the height difference Δ h1, Δ h2 of the ground point p3 and the non-ground points p4, p5 divided by the horizontal distances Δ d1, Δ d2 therebetween. Here it can be seen that the slope k1 between the ground point p3 and the non-ground point p4 is less than the slope threshold k0, while the slope k2 between the ground point p3 and the non-ground point p5 is greater than the slope threshold k 0. This means that point p4 belongs to and is added to the ground points as a false negative error (e.g., due to uneven or sloping road surfaces) is generated in the ground model estimation, while point p5 belongs to a correctly classified point and thus continues to remain as a non-ground point.
FIG. 4 shows a block diagram of an apparatus for ground filtering based on a laser point cloud according to an exemplary embodiment of the present invention.
The apparatus 400 includes an acquisition module 401 to acquire laser point cloud data. The acquisition module 401 may be configured directly as a lidar sensor and may generate a laser point cloud, for example, by scanning the surroundings. Optionally, the acquisition module 401 may also be configured as an interface and receive laser point cloud data from a lidar sensor. Illustratively, the acquisition module 401 may also include a selection device (not shown) and a subset of the laser point cloud data can be selected by the selection device for further processing.
The acquisition module 401 is coupled with the estimation module 402 to provide the acquired laser point cloud data to the estimation module 401. In the estimation module 402, a ground model can be estimated from the laser point cloud data, for example, based on model fitting.
Furthermore, the apparatus 400 comprises a verification module 403 configured to verify the estimated ground model based on the feature calculation and comparison.
Fig. 5 shows a block diagram of a lidar system according to an exemplary embodiment of the invention.
For further analysis processing of the laser point cloud data, the resulting laser point cloud data is provided by a signal processing apparatus 503 to an apparatus 504. Device 504 may then perform corresponding ground filtering based on the laser point cloud data.
Although specific embodiments of the invention have been described herein in detail, they have been presented for purposes of illustration only and are not to be construed as limiting the scope of the invention. Various substitutions, alterations, and modifications may be devised without departing from the spirit and scope of the present invention.
Claims (10)
1. A method for ground filtering based on a laser point cloud, the method comprising the steps of:
s1: acquiring laser point cloud data;
s2: estimating a ground model from the laser point cloud data based on model fitting; and
s3: and verifying the estimated ground model based on the feature comparison.
2. The method according to claim 1, wherein, prior to step S2, additionally performing:
the complete laser point cloud is divided, in particular non-uniformly divided, into a plurality of grid regions according to the region positions and sizes in the laser point cloud.
3. The method of claim 2, wherein the method further comprises:
the size of the grid area is determined in dependence on the distance to the lidar sensor, preferably the size of the grid area is increased with increasing distance to the lidar sensor.
4. The method according to any of the preceding claims, wherein said step S2 comprises:
selecting a subset of laser point cloud data;
establishing an initial ground model based on points in the subset of the laser point cloud data; and
comparing distances of further points in the laser point cloud data with respect to the initial ground model to a predefined distance threshold, wherein points whose distances with respect to the initial ground model are below the distance threshold are classified as ground points and points whose distances with respect to the initial ground model exceed the distance threshold are classified as non-ground points.
5. The method according to any of the preceding claims, wherein said step S2 comprises: random sampling fitting is carried out on the laser point cloud data by means of a RANSAC algorithm, and an optimal ground model is estimated through iterative calculation.
6. The method according to any of the preceding claims, wherein said step S3 comprises:
calculating a feature vector for each point in the laser point cloud data;
calculating a reference vector for the ground model estimated in step S2;
comparing the feature vector of each point with the reference vector; and
false positive errors and false negative errors in the ground model estimate are rejected based on the results of the comparison.
7. The method of claim 6, wherein the feature vector comprises a normal vector, a directional gradient vector, and/or a height gradient vector.
8. An apparatus (400, 504) for ground filtering based on a laser point cloud, the apparatus being configured to perform the method according to any one of claims 1 to 7, the apparatus (400, 504) comprising:
an acquisition module (401) configured to be able to acquire laser point cloud data;
an estimation module (402) configured to estimate a ground model from the laser point cloud data based on a model fit; and
a verification module (403) configured to enable verification of the estimated ground model based on the feature comparison.
9. A lidar system (500), the lidar system (500) comprising:
a lidar sensor (502) configured to be capable of transmitting and receiving radar signals;
a signal processing device (503) configured to be able to process lidar signals of the lidar sensor (502) and for outputting laser point cloud data; and
the apparatus (400, 504) for ground filtering based on laser point cloud of claim 8.
10. A machine readable program carrier having stored thereon a computer program for performing the method according to any one of claims 1 to 7 when the computer program runs on a computer.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202011559390.8A CN112669458A (en) | 2020-12-25 | 2020-12-25 | Method, device and program carrier for ground filtering based on laser point cloud |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202011559390.8A CN112669458A (en) | 2020-12-25 | 2020-12-25 | Method, device and program carrier for ground filtering based on laser point cloud |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN112669458A true CN112669458A (en) | 2021-04-16 |
Family
ID=75408818
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202011559390.8A Pending CN112669458A (en) | 2020-12-25 | 2020-12-25 | Method, device and program carrier for ground filtering based on laser point cloud |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN112669458A (en) |
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN113713381A (en) * | 2021-09-09 | 2021-11-30 | 腾讯科技（深圳）有限公司 | Object management method, device, equipment, storage medium and system |
| CN113759947A (en) * | 2021-09-10 | 2021-12-07 | 中航空管系统装备有限公司 | Airborne flight obstacle avoidance auxiliary method, device and system based on laser radar |
| CN114119998A (en) * | 2021-12-01 | 2022-03-01 | 成都理工大学 | Vehicle-mounted point cloud ground point extraction method and storage medium |
-
2020
- 2020-12-25 CN CN202011559390.8A patent/CN112669458A/en active Pending
Cited By (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN113713381A (en) * | 2021-09-09 | 2021-11-30 | 腾讯科技（深圳）有限公司 | Object management method, device, equipment, storage medium and system |
| CN113713381B (en) * | 2021-09-09 | 2023-06-20 | 腾讯科技（深圳）有限公司 | Object management method, device, equipment, storage medium and system |
| CN113759947A (en) * | 2021-09-10 | 2021-12-07 | 中航空管系统装备有限公司 | Airborne flight obstacle avoidance auxiliary method, device and system based on laser radar |
| CN113759947B (en) * | 2021-09-10 | 2023-08-08 | 中航空管系统装备有限公司 | Airborne flight obstacle avoidance assisting method, device and system based on laser radar |
| CN114119998A (en) * | 2021-12-01 | 2022-03-01 | 成都理工大学 | Vehicle-mounted point cloud ground point extraction method and storage medium |
| CN114119998B (en) * | 2021-12-01 | 2023-04-18 | 成都理工大学 | Vehicle-mounted point cloud ground point extraction method and storage medium |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN111882612B (en) | Vehicle multi-scale positioning method based on three-dimensional laser detection lane line | |
| CN108319655B (en) | Method and device for generating grid map | |
| US11222204B2 (en) | Creation of a 3D city model from oblique imaging and lidar data | |
| CN112669458A (en) | Method, device and program carrier for ground filtering based on laser point cloud | |
| CN108868268B (en) | Unmanned parking space posture estimation method based on point-to-surface distance and cross-correlation entropy registration | |
| CN109598765B (en) | Monocular camera and millimeter wave radar external parameter combined calibration method based on spherical calibration object | |
| Kang et al. | Automatic targetless camera–lidar calibration by aligning edge with gaussian mixture model | |
| Huber et al. | A new approach to 3-d terrain mapping | |
| Xiao et al. | Planar segment based three‐dimensional point cloud registration in outdoor environments | |
| Hebel et al. | Simultaneous calibration of ALS systems and alignment of multiview LiDAR scans of urban areas | |
| CN108828621A (en) | Obstacle detection and road surface partitioning algorithm based on three-dimensional laser radar | |
| CN112464812B (en) | Vehicle-based concave obstacle detection method | |
| WO2022188663A1 (en) | Target detection method and apparatus | |
| CN111273312B (en) | Intelligent vehicle positioning and loop detection method | |
| Zhang et al. | Three-dimensional cooperative mapping for connected and automated vehicles | |
| CN111457930B (en) | High-precision mapping positioning method by combining vehicle-mounted Lidar and unmanned aerial vehicle | |
| CN112381026A (en) | Method, system and device for detecting channel shoreline and storage medium | |
| CN114051628A (en) | Method and device for determining target object point cloud set | |
| AU2022259832A1 (en) | Target detection in a point cloud | |
| CN112102375A (en) | Method and device for detecting reliability of point cloud registration and mobile intelligent equipment | |
| CN113835078A (en) | Signal level joint detection method and device based on local three-dimensional grid | |
| CN117392237A (en) | Robust laser radar-camera self-calibration method | |
| CN113536959A (en) | Dynamic obstacle detection method based on stereoscopic vision | |
| CN110554405B (en) | Normal scanning registration method and system based on cluster combination | |
| Hebel et al. | Pre-classification of points and segmentation of urban objects by scan line analysis of airborne LiDAR data |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication |