CN114140470A - Ground object semantic segmentation method based on helicopter airborne laser radar - Google Patents
Ground object semantic segmentation method based on helicopter airborne laser radar Download PDFInfo
- Publication number
- CN114140470A CN114140470A CN202111484048.0A CN202111484048A CN114140470A CN 114140470 A CN114140470 A CN 114140470A CN 202111484048 A CN202111484048 A CN 202111484048A CN 114140470 A CN114140470 A CN 114140470A
- Authority
- CN
- China
- Prior art keywords
- point cloud
- dimensional
- helicopter
- laser radar
- semantic segmentation
- 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
- 238000000034 method Methods 0.000 title claims abstract description 37
- 230000011218 segmentation Effects 0.000 title claims abstract description 27
- 238000013135 deep learning Methods 0.000 claims abstract description 14
- 238000005070 sampling Methods 0.000 claims abstract description 7
- 238000007781 pre-processing Methods 0.000 claims abstract description 4
- 238000010586 diagram Methods 0.000 claims description 32
- 238000000605 extraction Methods 0.000 claims description 4
- 230000006870 function Effects 0.000 claims description 4
- 238000006243 chemical reaction Methods 0.000 claims description 3
- 238000007635 classification algorithm Methods 0.000 claims description 3
- 238000001914 filtration Methods 0.000 claims description 3
- 238000012545 processing Methods 0.000 abstract description 10
- 238000001514 detection method Methods 0.000 abstract description 9
- 238000000926 separation method Methods 0.000 description 7
- 230000000007 visual effect Effects 0.000 description 4
- 238000004422 calculation algorithm Methods 0.000 description 3
- 238000005259 measurement Methods 0.000 description 2
- 238000011160 research Methods 0.000 description 2
- 238000006467 substitution reaction Methods 0.000 description 2
- 238000013473 artificial intelligence Methods 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 238000004364 calculation method Methods 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 230000018109 developmental process Effects 0.000 description 1
- 238000005457 optimization Methods 0.000 description 1
- 230000008447 perception Effects 0.000 description 1
- 230000000717 retained effect Effects 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/10—Segmentation; Edge detection
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/23—Clustering techniques
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
- G06F18/241—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/045—Combinations of networks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
-
- 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
-
- 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
- G06T2207/10032—Satellite or aerial image; Remote sensing
- G06T2207/10044—Radar image
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Data Mining & Analysis (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- Life Sciences & Earth Sciences (AREA)
- Artificial Intelligence (AREA)
- Evolutionary Computation (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Bioinformatics & Computational Biology (AREA)
- Evolutionary Biology (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Health & Medical Sciences (AREA)
- Biomedical Technology (AREA)
- Biophysics (AREA)
- Computational Linguistics (AREA)
- General Health & Medical Sciences (AREA)
- Molecular Biology (AREA)
- Computing Systems (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Optical Radar Systems And Details Thereof (AREA)
Abstract
The invention relates to a ground object semantic segmentation method based on a helicopter airborne laser radar, which relates to the technical field of real-time processing of the helicopter airborne laser radar and comprises the following steps: the method comprises the steps of obtaining real-time point cloud through a laser radar, collecting point cloud data, enabling the collected point cloud data to enter a down-sampling module, reducing the data volume of the point cloud while keeping the characteristics of the point cloud, obtaining effective data, carrying out preprocessing on the effective data obtained after down-sampling to encode three-dimensional point cloud into a two-dimensional aerial view, dividing points into places and object points in a deep learning network mode, carrying out obstacle detection only in the object points, dividing an object point set into a plurality of isolated point cloud clusters, classifying by using the characteristics of the point cloud clusters, calculating external polyhedrons of the clustered point cloud belonging to the same target, and distinguishing objects to be detected. The helicopter flight safety detection device has the advantages that the requirements of real-time performance and low power consumption can be met, the detection performance is better, and the safety requirement of helicopter flight is guaranteed.
Description
Technical Field
The invention relates to the technical field of real-time processing of helicopter airborne laser radars, in particular to a ground object semantic segmentation method based on the helicopter airborne laser radars.
Background
The laser radar has the advantages of high measurement precision, fine time and space resolution, long measurement distance and the like, the helicopter can fly and lift, hover, pitch, deflect and other various complex actions with low altitude, low speed and unchanged nose direction, has the characteristics of flexibility and is very suitable for measuring various complex terrains. Therefore, helicopter-mounted laser radars are increasingly widely applied in many aspects such as geodetic surveying, forest exploration, urban modeling, disaster assessment and the like.
With the rapid development of artificial intelligence and the field of helicopters, scene understanding is crucial to the safety and effectiveness of machine automatic perception in complex dynamic scenes. Various sensors are generally equipped in helicopters, and particularly lidar sensors play an important role in understanding the visual environment, and lidar systems are used for collecting sparse 3D point clouds to reconstruct the environment in the actual scene and help automatic systems make decisions to better understand the scene, so scene semantic understanding of the point clouds is crucial to helicopter flight; meanwhile, researches find that the ground can provide useful information, ambiguity caused by data sparsity is effectively eliminated, and the relation between an object and the ground is beneficial to semantic segmentation and prediction, so that the important points of the researches are how to effectively segment point cloud semantics of a helicopter flight scene and how to segment the ground and obtain the relation between the object and the ground.
The existing methods are classified into two types based on deep learning and non-deep learning, and the method based on deep learning has high calculation force demand and is difficult to simultaneously meet the requirements of real-time performance and low power consumption under the condition of an airborne vehicle; the method based on non-deep learning has poor detection performance and is difficult to ensure the flight safety of the helicopter.
The foregoing description is provided for general background information and is not admitted to be prior art.
Disclosure of Invention
The invention aims to provide a ground object semantic segmentation method based on an airborne laser radar of a helicopter, which can meet the requirements of real-time performance and low power consumption at the same time, has better detection performance and ensures the safety requirement of the flight of the helicopter.
The invention provides a ground object semantic segmentation method based on a helicopter airborne laser radar, which is characterized by comprising the following steps of:
s1: acquiring real-time point cloud through a laser radar, collecting point cloud data, and performing S2;
s2: the collected point cloud data enters a down-sampling module, the point cloud characteristics are kept, meanwhile, the point cloud data amount is reduced, effective data are obtained, and the step S3 is carried out;
s3: preprocessing the effective data obtained after the downsampling to encode the three-dimensional point cloud into a two-dimensional aerial view, and entering the step S4;
s4: dividing the points into places and object points by using a deep learning network mode, detecting obstacles only in the object points, and entering the step S5;
s5: dividing the object point set into a plurality of isolated point cloud clusters by using an Euclidean clustering mode, and entering the step S6;
s6: and classifying by using the characteristics of the point cloud clusters, calculating external polyhedrons of the clustered point clouds belonging to the same target, and classifying the external polyhedrons according to the attributes of the external polyhedrons to obtain the objects to be detected.
Further, step S1 includes the steps of:
s11: sending a udp packet by the laser radar;
s12: the server end receives and unpacks the packet, and eliminates the dead points to perform coordinate system conversion;
s13: and finishing the acquisition of point cloud data.
Further, down-sampling is performed by voxel filtering in step S2.
Further, step S3 includes the steps of:
s31: encoding the three-dimensional point cloud into a two-dimensional aerial view within the roi range;
s32: and encoding the two-dimensional feature map by using the highest height value of the point clouds falling into the same grid and the median of the height of the point clouds falling into the same grid.
Further, the object to be detected comprises a building, a rod, a windmill, a high-voltage line tower and a signal tower.
Further, the step S3 further includes:
establishing an roi, wherein x is-128 m < x is less than or equal to 128m, z is 0< z is less than or equal to 256m in a laser radar coordinate system, converting the point cloud into a pixel coordinate system, counting by using a mode that the highest point and a median of the point cloud height falling into the same grid replace the value of the pixel according to a 0.5mX0.5m space, wherein the converted three-channel value is (x, y, z), and the converted five-channel value is (x, y, z, yaw, pitch), so that a two-dimensional characteristic diagram of five channels and one three channel is obtained and used as the input of the network.
Further, the yaw and the pitch are two angles that can be obtained in three dimensions, respectively.
Further, taking the two-dimensional feature map of the five channels in the step S3 as an input of the network, and obtaining a feature map by using a 1-by-1 convolution kernel with an output channel of 32 channels; and outputting a prediction species map by using a 32-channel point cloud three-dimensional coordinate tensor characteristic map through multiple characteristic extraction, multiple characteristic map splicing, multiple upsampling and loss function input, and obtaining a point cloud map by using a 1-channel point cloud three-dimensional coordinate tensor characteristic map through a nearest neighbor classification algorithm random condition field.
Further, the backbone specifically comprises the following steps:
and (3) performing three times of filling convolution on the three-channel two-dimensional characteristic diagram in the step (S3), dividing the three-channel two-dimensional characteristic diagram into a characteristic diagram with the channel number 9 times of the original channel number by the convolution kernel size of (3, 3) of the current backbone input characteristic diagram, multiplying the characteristic diagram by two points, obtaining the convolution result of the channel number of the input characteristic diagram by the two-dimensional convolution of (1, 1), and splicing the convolution result with the characteristic diagram of the original backbone input characteristic diagram by the two-dimensional convolution of (3, 3) to complete the backbone.
According to the ground object semantic segmentation method based on the helicopter-borne laser radar, the data amount required to be processed can be greatly reduced through ground object separation, the difficulty of subsequent target identification can be greatly reduced through accurate ground object separation, and therefore a deep learning method is required to be used; subsequent external polyhedron estimation and target classification can be carried out by using a non-deep learning method under the condition of limited computing resources, so that the real-time performance is ensured; the encoding method is different from the existing method, three-dimensional point cloud is encoded into a two-dimensional aerial view in the roi range, the two-dimensional feature map is encoded by using the highest height value (radar information z) of the point cloud falling into the same grid and the median of the point cloud height falling into the same grid, stable point cloud density can be kept in a range far away from the point cloud, and the requirement of a helicopter on the detection precision of far-end interest is met; in consideration of the characteristic of large visual field range under an airborne condition, structures for improving the visual field range, such as u-shaped structures and the like, are added in the original squeezeseg model.
Drawings
Fig. 1 is an algorithm block diagram of a ground object semantic segmentation method based on a helicopter airborne laser radar according to an embodiment of the present invention.
Fig. 2 is a flowchart of the ground object semantic segmentation method based on the helicopter airborne laser radar in fig. 1.
Fig. 3 is a processing module diagram of step S4 of the semantic segmentation method for ground targets based on laser radar on board the helicopter in fig. 1.
Fig. 4 is an overall processing module diagram of the ground object semantic segmentation method based on the helicopter airborne laser radar in fig. 1.
Detailed Description
The following detailed description of embodiments of the present invention is provided in connection with the accompanying drawings and examples. The following examples are intended to illustrate the invention but are not intended to limit the scope of the invention.
The terms first, second, third, fourth and the like in the description and in the claims of the present invention are used for distinguishing between similar elements and not necessarily for describing a particular sequential or chronological order.
Example 1
Fig. 1 is an algorithm block diagram of a ground object semantic segmentation method based on a helicopter airborne laser radar according to an embodiment of the present invention. Referring to fig. 1, the method for semantic segmentation of ground objects based on an airborne laser radar of a helicopter provided by the embodiment of the present invention includes the following steps:
s1: acquiring real-time point cloud through a laser radar, collecting point cloud data, and performing S2;
s2: the collected point cloud data enters a down-sampling module, the point cloud characteristics are kept, meanwhile, the point cloud data amount is reduced, effective data are obtained, it is required to be noted that the processing time can be reduced through the down-sampling module, and the step S3 is carried out;
s3: preprocessing the effective data obtained after the downsampling to encode the three-dimensional point cloud into a two-dimensional aerial view, and entering the step S4;
s4: dividing the points into places and object points (points not in the earth) by using a deep learning network mode, detecting obstacles only in the object points, and entering the step S5;
s5: dividing the object point set into a plurality of isolated point cloud clusters by using an Euclidean clustering mode (clustering non-ground target point clouds), and entering the step S6;
s6: and classifying by using the characteristics of the point cloud clusters, calculating external polyhedrons of the clustered point clouds belonging to the same target, and classifying the external polyhedrons according to the attributes of the external polyhedrons to obtain the objects to be detected. Specifically, the object to be detected includes a building, a shaft, a windmill, a high-voltage line tower, a signal tower, and the like.
According to the ground object semantic segmentation method based on the helicopter-borne laser radar, the data amount required to be processed can be greatly reduced through ground object separation, the difficulty of subsequent target identification can be greatly reduced through accurate ground object separation, and therefore a deep learning method is required to be used; subsequent external polyhedron estimation and target classification can be carried out by using a non-deep learning method under the condition of limited computing resources, so that the real-time performance is ensured; under the condition that computing resources are limited, the deep learning network is not suitable for using a point cloud network with overlarge computing amount, and optimization is carried out based on the image network, so that point cloud data can be processed, and instantaneity is guaranteed.
Fig. 2 is a flowchart of the ground object semantic segmentation method based on the helicopter airborne laser radar in fig. 1. Referring to fig. 2, step S1 includes the following steps:
s11: the laser radar sends udp (user data packet protocol) packets;
s12: a server (receiving) end receives and unpacks a packet, and eliminates a dead point to carry out coordinate system conversion;
specifically, the dead spots include points where the distance or the reflection intensity is not reasonable and invalid points.
S13: and finishing the acquisition of point cloud data.
Further, in step S2, downsampling is performed through voxel filtering, so that the point cloud data amount is reduced while the point cloud features are effectively retained, and the processing time is reduced.
Step S3 includes the following steps:
s31: encoding the three-dimensional point cloud into a two-dimensional aerial view within the roi (region of interest);
s32: and encoding the two-dimensional feature map by using the highest height value of the point clouds falling into the same grid and the median of the height of the point clouds falling into the same grid.
It should be noted that, the foresight maps convert point clouds into two-dimensional feature maps:
firstly, establishing a roi (region of interest), for example, in a laser radar coordinate system, 128m < x < 128m < 0< z < 256m, only converting the point cloud in the range, converting the point cloud into a pixel coordinate system, and having a statistical value generation process, wherein a statistical unit (pixel of a reference picture) corresponds to a 0.5mx0.5m space, and the value of the pixel is replaced by a highest point (radar information z is minimum) and a median of the point cloud height falling into the same grid (three channels of an analog color picture are values of a color space, such as blue, green, and red components in a bgr color space), the converted three channel values are (x, y, z), the converted five channel values are (x, y, z, yaw, pitch), yaw and pitch are two three-dimensional calculated angles; two-dimensional characteristic maps of one five channel and one three channel are obtained as the input of the network.
Fig. 4 is an overall processing module diagram of the ground object semantic segmentation method based on the helicopter airborne laser radar in fig. 1. Referring to fig. 4, the two-dimensional feature map of the five channels in step S3 is used as an input of the network, and the output channel is 32 channels by a 1-by-1 convolution kernel to obtain a feature map; and outputting a prediction species map by using a 32-channel point cloud three-dimensional coordinate tensor characteristic map through multiple characteristic extraction, multiple characteristic map splicing, multiple upsampling and loss function input, and obtaining a point cloud map by using a 1-channel point cloud three-dimensional coordinate tensor characteristic map through a nearest neighbor classification algorithm random condition field.
Fig. 3 is a processing module diagram of step S4 of the semantic segmentation method for ground targets based on laser radar on board the helicopter in fig. 1. Referring to fig. 3, the background (feature extraction) specifically includes the following steps:
and (3) performing three times of filling convolution on the three-channel two-dimensional characteristic diagram in the step (S3), dividing the three-channel two-dimensional characteristic diagram into a characteristic diagram with the channel number 9 times of the original channel number by the convolution kernel size of (3, 3) of the current backbone input characteristic diagram, multiplying the characteristic diagram by two points, obtaining the convolution result of the channel number of the input characteristic diagram by the two-dimensional convolution of (1, 1), and splicing the convolution result with the characteristic diagram of the original backbone input characteristic diagram by the two-dimensional convolution of (3, 3) to complete the backbone.
Based on the above description, the present invention has the following advantages:
1. through a semantic segmentation algorithm based on deep learning, the ground feature separation function of laser point cloud data is realized, the data volume of subsequent processing can be reduced, the reliability of the subsequent processing such as target identification can be improved, and the accuracy and the robustness of ground feature separation can be improved compared with a common ground feature separation method based on a priori map;
2. considering the requirements of large laser radar visual field range and detection of small targets such as cables, people and the like under an airborne condition, a U-shaped structure is newly added into an original SqueezeSeg network, so that the detail information of a deep structure of the network is enriched, and the detection performance of the small targets is improved;
3. when the laser point cloud is subjected to two-dimensional coding, a conventional coding mode based on angular resolution is abandoned, a coding mode based on distance is adopted, and the point cloud is less when the distance is far, and the helicopter focuses more on the detection performance of a remote target.
The above description is only for the specific embodiments of the present invention, but the scope of the present invention is not limited thereto, and any person skilled in the art can easily conceive of the changes or substitutions within the technical scope of the present invention, and all the changes or substitutions should be covered within the scope of the present invention. Therefore, the protection scope of the present invention shall be subject to the protection scope of the appended claims.
Claims (9)
1. A ground object semantic segmentation method based on a helicopter airborne laser radar is characterized by comprising the following steps:
s1: acquiring real-time point cloud through a laser radar, collecting point cloud data, and performing S2;
s2: the collected point cloud data enters a down-sampling module, the point cloud characteristics are kept, meanwhile, the point cloud data amount is reduced, effective data are obtained, and the step S3 is carried out;
s3: preprocessing the effective data obtained after the downsampling to encode the three-dimensional point cloud into a two-dimensional aerial view, and entering the step S4;
s4: dividing the points into places and object points by using a deep learning network mode, detecting obstacles only in the object points, and entering the step S5;
s5: dividing the object point set into a plurality of isolated point cloud clusters by using an Euclidean clustering mode, and entering the step S6;
s6: and classifying by using the characteristics of the point cloud clusters, calculating external polyhedrons of the clustered point clouds belonging to the same target, and classifying the external polyhedrons according to the attributes of the external polyhedrons to obtain the objects to be detected.
2. The method for semantic segmentation of ground targets based on airborne lidar of a helicopter according to claim 1, wherein step S1 comprises the steps of:
s11: sending a udp packet by the laser radar;
s12: the server end receives and unpacks the packet, and eliminates the dead points to perform coordinate system conversion;
s13: and finishing the acquisition of point cloud data.
3. The helicopter airborne lidar based ground target semantic segmentation method of claim 1, wherein the downsampling is performed by voxel filtering in step S2.
4. The method for semantic segmentation of ground targets based on airborne lidar of a helicopter according to claim 1, wherein step S3 comprises the steps of:
s31: encoding the three-dimensional point cloud into a two-dimensional aerial view within the roi range;
s32: and encoding the two-dimensional feature map by using the highest height value of the point clouds falling into the same grid and the median of the height of the point clouds falling into the same grid.
5. The method for semantic segmentation of ground targets based on helicopter-borne laser radar according to claim 1, characterized in that the objects to be detected comprise buildings, shafts, windmills, high-voltage line towers, signal towers.
6. The method for semantic segmentation of ground targets based on airborne lidar of a helicopter according to claim 1, wherein said step S3 further comprises:
establishing an roi, wherein x is-128 m < x is less than or equal to 128m, z is 0< z is less than or equal to 256m in a laser radar coordinate system, converting the point cloud into a pixel coordinate system, counting by using a mode that the highest point and a median of the point cloud height falling into the same grid replace the value of the pixel according to a 0.5mX0.5m space, wherein the converted three-channel value is (x, y, z), and the converted five-channel value is (x, y, z, yaw, pitch), so that a two-dimensional characteristic diagram of five channels and one three channel is obtained and used as the input of the network.
7. The method according to claim 6, wherein yaw and pitch are two angles that can be determined in three dimensions.
8. The method for semantic segmentation of ground targets based on airborne lidar of a helicopter according to claim 6, characterized in that the five-channel two-dimensional feature map of step S3 is used as input of a network, and a feature map is obtained by a 1-by-1 convolution kernel with 32 output channels; and outputting a prediction species map by using a 32-channel point cloud three-dimensional coordinate tensor characteristic map through multiple characteristic extraction, multiple characteristic map splicing, multiple upsampling and loss function input, and obtaining a point cloud map by using a 1-channel point cloud three-dimensional coordinate tensor characteristic map through a nearest neighbor classification algorithm random condition field.
9. The helicopter airborne lidar based ground target semantic segmentation method of claim 8, wherein the backhaul specifically comprises the steps of:
and (3) performing three times of filling convolution on the three-channel two-dimensional characteristic diagram in the step (S3), dividing the three-channel two-dimensional characteristic diagram into a characteristic diagram with the channel number 9 times of the original channel number by the convolution kernel size of (3, 3) of the current backbone input characteristic diagram, multiplying the characteristic diagram by two points, obtaining the convolution result of the channel number of the input characteristic diagram by the two-dimensional convolution of (1, 1), and splicing the convolution result with the characteristic diagram of the original backbone input characteristic diagram by the two-dimensional convolution of (3, 3) to complete the backbone.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202111484048.0A CN114140470A (en) | 2021-12-07 | 2021-12-07 | Ground object semantic segmentation method based on helicopter airborne laser radar |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202111484048.0A CN114140470A (en) | 2021-12-07 | 2021-12-07 | Ground object semantic segmentation method based on helicopter airborne laser radar |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN114140470A true CN114140470A (en) | 2022-03-04 |
Family
ID=80384783
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202111484048.0A Pending CN114140470A (en) | 2021-12-07 | 2021-12-07 | Ground object semantic segmentation method based on helicopter airborne laser radar |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN114140470A (en) |
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN115356740A (en) * | 2022-08-09 | 2022-11-18 | 群周科技(上海)有限公司 | Landing positioning method for landing area in airborne environment |
| WO2023193400A1 (en) * | 2022-04-06 | 2023-10-12 | 合众新能源汽车股份有限公司 | Point cloud detection and segmentation method and apparatus, and electronic device |
| CN117292140A (en) * | 2023-10-17 | 2023-12-26 | 小米汽车科技有限公司 | Point cloud data processing method and device, vehicle and storage medium |
Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111507982A (en) * | 2019-06-28 | 2020-08-07 | 浙江大学 | Point cloud semantic segmentation method based on deep learning |
| CN111583337A (en) * | 2020-04-25 | 2020-08-25 | 华南理工大学 | Omnibearing obstacle detection method based on multi-sensor fusion |
| CN112200248A (en) * | 2020-10-13 | 2021-01-08 | 北京理工大学 | Point cloud semantic segmentation method, system and storage medium under urban road environment based on DBSCAN clustering |
| WO2021004813A1 (en) * | 2019-07-08 | 2021-01-14 | Continental Automotive Gmbh | Method and mobile entity for detecting feature points in an image |
| US10929694B1 (en) * | 2020-01-22 | 2021-02-23 | Tsinghua University | Lane detection method and system based on vision and lidar multi-level fusion |
-
2021
- 2021-12-07 CN CN202111484048.0A patent/CN114140470A/en active Pending
Patent Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111507982A (en) * | 2019-06-28 | 2020-08-07 | 浙江大学 | Point cloud semantic segmentation method based on deep learning |
| WO2021004813A1 (en) * | 2019-07-08 | 2021-01-14 | Continental Automotive Gmbh | Method and mobile entity for detecting feature points in an image |
| US10929694B1 (en) * | 2020-01-22 | 2021-02-23 | Tsinghua University | Lane detection method and system based on vision and lidar multi-level fusion |
| CN111583337A (en) * | 2020-04-25 | 2020-08-25 | 华南理工大学 | Omnibearing obstacle detection method based on multi-sensor fusion |
| CN112200248A (en) * | 2020-10-13 | 2021-01-08 | 北京理工大学 | Point cloud semantic segmentation method, system and storage medium under urban road environment based on DBSCAN clustering |
Non-Patent Citations (1)
| Title |
|---|
| CHENFENG XU, BICHEN WU, ZINING WANG, WEI ZHAN, PETER VAJDA, KURT KEUTZER, AND MASAYOSHI TOMIZUKA: "SqueezeSegV3: Spatially-Adaptive Convolution for Efficient Point-Cloud Segmentation", COMPUTER VISION -- ECCV 2020 : PART XXVIII, 13 April 2021 (2021-04-13), pages 1 - 19 * |
Cited By (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| WO2023193400A1 (en) * | 2022-04-06 | 2023-10-12 | 合众新能源汽车股份有限公司 | Point cloud detection and segmentation method and apparatus, and electronic device |
| CN115356740A (en) * | 2022-08-09 | 2022-11-18 | 群周科技(上海)有限公司 | Landing positioning method for landing area in airborne environment |
| CN117292140A (en) * | 2023-10-17 | 2023-12-26 | 小米汽车科技有限公司 | Point cloud data processing method and device, vehicle and storage medium |
| CN117292140B (en) * | 2023-10-17 | 2024-04-02 | 小米汽车科技有限公司 | Point cloud data processing method and device, vehicle and storage medium |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN111832655B (en) | Multi-scale three-dimensional target detection method based on characteristic pyramid network | |
| CN114140470A (en) | Ground object semantic segmentation method based on helicopter airborne laser radar | |
| Chen et al. | Distribution line pole detection and counting based on YOLO using UAV inspection line video | |
| CN113902897B (en) | Training of target detection model, target detection method, device, equipment and medium | |
| CN113610044B (en) | 4D millimeter wave three-dimensional target detection method and system based on self-attention mechanism | |
| CN106094569A (en) | Multi-sensor Fusion unmanned plane perception with evade analogue system and emulation mode thereof | |
| CN111126184B (en) | Post-earthquake building damage detection method based on unmanned aerial vehicle video | |
| CN111008975B (en) | Mixed pixel unmixing method and system for space artificial target linear model | |
| CN107479065B (en) | Forest gap three-dimensional structure measuring method based on laser radar | |
| CN114140471A (en) | Ground target semantic segmentation method based on helicopter airborne laser radar | |
| CN112288667B (en) | Three-dimensional target detection method based on fusion of laser radar and camera | |
| CN110717496B (en) | Complex scene tree detection method based on neural network | |
| Awrangjeb et al. | Rule-based segmentation of LIDAR point cloud for automatic extraction of building roof planes | |
| CN113269147B (en) | Three-dimensional detection method and system based on space and shape, and storage and processing device | |
| CN117274749B (en) | Fused 3D target detection method based on 4D millimeter wave radar and image | |
| CN112630160A (en) | Unmanned aerial vehicle track planning soil humidity monitoring method and system based on image acquisition and readable storage medium | |
| CN114038193A (en) | Intelligent traffic flow data statistical method and system based on unmanned aerial vehicle and multi-target tracking | |
| CN117808689A (en) | Depth complement method based on fusion of millimeter wave radar and camera | |
| CN111458691B (en) | Building information extraction method and device and computer equipment | |
| CN114359754B (en) | Unmanned aerial vehicle power inspection laser point cloud real-time power transmission wire extraction method | |
| CN111337932A (en) | Power grid infrastructure construction checking method based on airborne laser radar system | |
| Li et al. | Vehicle object detection based on rgb-camera and radar sensor fusion | |
| CN111812670B (en) | Single photon laser radar space transformation noise judgment and filtering method and device | |
| CN112613437B (en) | Method for identifying illegal buildings | |
| WO2021179583A1 (en) | Detection method and detection device |
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 |