CN110674829B - Three-dimensional target detection method based on graph convolution attention network - Google Patents
Three-dimensional target detection method based on graph convolution attention network Download PDFInfo
- Publication number
- CN110674829B CN110674829B CN201910918980.6A CN201910918980A CN110674829B CN 110674829 B CN110674829 B CN 110674829B CN 201910918980 A CN201910918980 A CN 201910918980A CN 110674829 B CN110674829 B CN 110674829B
- Authority
- CN
- China
- Prior art keywords
- convolution
- layer
- voxel
- point
- feature map
- 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
- 238000001514 detection method Methods 0.000 title claims abstract description 18
- 238000000605 extraction Methods 0.000 claims abstract description 19
- 238000000034 method Methods 0.000 claims description 15
- 230000006870 function Effects 0.000 claims description 7
- 238000013507 mapping Methods 0.000 claims description 6
- 230000008447 perception Effects 0.000 claims description 3
- 230000007246 mechanism Effects 0.000 abstract description 13
- 238000000638 solvent extraction Methods 0.000 abstract 1
- 230000008569 process Effects 0.000 description 3
- 238000010586 diagram Methods 0.000 description 2
- 230000002708 enhancing effect Effects 0.000 description 2
- 238000012545 processing Methods 0.000 description 2
- 208000000044 Amnesia Diseases 0.000 description 1
- 208000026139 Memory disease Diseases 0.000 description 1
- 238000013459 approach Methods 0.000 description 1
- 238000013528 artificial neural network Methods 0.000 description 1
- 230000003190 augmentative effect Effects 0.000 description 1
- 238000004364 calculation method Methods 0.000 description 1
- 230000001149 cognitive effect Effects 0.000 description 1
- 230000002860 competitive effect Effects 0.000 description 1
- 238000005520 cutting process Methods 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 238000009826 distribution Methods 0.000 description 1
- 230000000694 effects Effects 0.000 description 1
- 230000006984 memory degeneration Effects 0.000 description 1
- 208000023060 memory loss Diseases 0.000 description 1
- 238000007500 overflow downdraw method Methods 0.000 description 1
- 238000003672 processing method Methods 0.000 description 1
- 238000005096 rolling process Methods 0.000 description 1
- 238000003860 storage Methods 0.000 description 1
- 238000012800 visualization Methods 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/40—Extraction of image or video features
- G06V10/44—Local feature extraction by analysis of parts of the pattern, e.g. by detecting edges, contours, loops, corners, strokes or intersections; Connectivity analysis, e.g. of connected components
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
- G06F18/213—Feature extraction, e.g. by transforming the feature space; Summarisation; Mappings, e.g. subspace methods
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
- G06F18/213—Feature extraction, e.g. by transforming the feature space; Summarisation; Mappings, e.g. subspace methods
- G06F18/2136—Feature extraction, e.g. by transforming the feature space; Summarisation; Mappings, e.g. subspace methods based on sparsity criteria, e.g. with an overcomplete basis
-
- 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
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V2201/00—Indexing scheme relating to image or video recognition or understanding
- G06V2201/07—Target detection
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V2201/00—Indexing scheme relating to image or video recognition or understanding
- G06V2201/08—Detecting or categorising vehicles
Landscapes
- Engineering & Computer Science (AREA)
- Data Mining & Analysis (AREA)
- Theoretical Computer Science (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Life Sciences & Earth Sciences (AREA)
- Evolutionary Biology (AREA)
- Evolutionary Computation (AREA)
- Bioinformatics & Computational Biology (AREA)
- General Engineering & Computer Science (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Artificial Intelligence (AREA)
- Multimedia (AREA)
- Image Analysis (AREA)
Abstract
The invention provides a three-dimensional target detection method based on a graph convolution attention network. (1) voxelized partitioning and random downsampling are carried out on point clouds; (2) performing local feature extraction in each grid voxel; (3) extracting a high-order feature map by middle layer convolution; (4) The area suggests the frame, category, and direction of the network predicted target. In order to enhance the connection relation between each point and adjacent points, the invention provides a feature extraction module which is based on an edge convolution form and introduces an attention mechanism, and meanwhile, a attention mechanism module with the same principle is introduced after an intermediate convolution layer, so that features of each channel of the feature map are reselected, and a more reasonable high-order feature map is obtained. The invention improves the target detection accuracy of the point cloud, and has good performance especially under the condition of serious shielding.
Description
Technical Field
The invention relates to a computer vision three-dimensional point cloud processing method, in particular to a three-dimensional target detection method.
Background
Object detection is a traditional visualization task that can identify and locate objects simultaneously, which is a prerequisite for achieving intelligent scenarios. The two-dimensional detection has reached unprecedented flourish today, but in fields such as mapping, indoor robot and augmented reality, three-dimensional detection is significantly better than two-dimensional. It can provide more position and attitude information and is one of the basic tasks of automatic driving environment perception. RGB images were once the dominant data format for the target detection task, but with the development of 3D sensors, lidar has become an increasingly popular detection tool in recent years.
Some laser radar and camera based methods now combine point cloud data with image data to achieve higher accuracy. However, the fusion method also faces the problem of excessive computational cost, so the single sensor method is still competitive. Many studies have shown that point clouds are a more appropriate form of data describing the shape of objects. The point cloud may better represent the euclidean distance and has no multi-scale issues. However, the point cloud is a sparse data, which makes the two-dimensional method difficult to apply directly.
In extracting features, most methods use a point-by-point processing manner and use a symmetric function to extract global features, and the concept ignores the connection and the relation between points. And compared with the picture data, the point cloud is a natural graph structure which is easy to construct links. Some studies use the idea of graph network, consider that the relationship between adjacent points and edges helps to enhance the expression of local features, and propose an edge convolution method. In the three-dimensional convolution, in consideration of the fact that in a defined voxel range, due to the sparsity of points, many voxels are empty, and the sparse convolution mode is used, so that the calculation speed can be improved and the video memory loss can be reduced while the convolution effect is not affected.
Disclosure of Invention
The invention aims to provide a three-dimensional target detection method based on a graph convolution attention network, which can improve the accuracy of point cloud target detection and has good performance under the condition of serious shielding.
The purpose of the invention is realized in the following way:
(1) Voxel division and random downsampling are carried out on the point cloud;
(2) Extracting local features from each grid voxel;
(3) Extracting a high-order feature map by middle layer convolution;
(4) The area suggests the frame, category, and direction of the network predicted target.
The invention may further include:
1. the voxel division and random downsampling of the point cloud specifically comprises the following steps: dividing an original point cloud by using the structure of the voxel grid, discarding outliers outside a specified range, dividing the point cloud into grids, randomly downsampling in each voxel grid, numbering each grid, and storing.
The storage is stored by using a hash table.
2. The local feature extraction in each grid voxel specifically comprises the following steps: within the grid of each voxel, a graph annotation network module is used to perform feature extraction on the corresponding points.
The feature extraction of the corresponding points by using the graph annotation network module is specifically as follows: the method comprises the steps of firstly connecting edges between each point and adjacent points around to form a graph structure taking Euclidean distance as a judgment standard, simultaneously connecting each point with the point, extracting information such as coordinates of two end points of each edge to serve as initial features of the edges, then carrying out convolution operation on the edges, and finally obtaining voxel level features through selection of a symmetric function.
Prior to the edge convolution operation, an attention mechanism is used to select the initial feature.
3. The middle layer convolution extraction high-order feature map specifically comprises the following steps: compressing the feature map into a compact structure by using a sparse convolution method, and mapping back to the original sparse space representation after convolution; after convolution abstraction, weight is redistributed to different channels by using an attention mechanism to obtain an attention map corresponding to the feature map, and the attention map is superimposed on the high-order feature map obtained by convolution to obtain a final three-dimensional feature map.
4. The frame, the category and the direction of the area suggestion network prediction target specifically comprise: and after the high-order feature map subjected to multi-layer convolution is subjected to feature extraction, calculating predicted values of boundary frames, categories and directions corresponding to each anchor point by using three respective full-connection layers.
The three-dimensional target detection method based on the graph convolution attention network is characterized by enhancing the process of expressing the local relationship of the point cloud and optimizing the feature selection. The invention uses the edge convolution method capable of expressing the relation between adjacent points for feature extraction of target detection, and uses a attention mechanism to select initial physical features which are more important for feature expression in the feature selection stage of initial points, thereby obtaining better extracted features. In the middle layer convolution process, multi-channel characteristic data are also generated, and the invention optimizes the convolution result by using the thought of a attention mechanism, strengthens the proportion of channels with main influence and obtains a characteristic diagram with more representation force.
The point cloud data of a typical set of scenes contains more than 100k points, so it is considered to preprocess, i.e. voxelise, the point cloud using a specific data structure. The original points are first divided into voxels and punctiform features are first extracted, then the down-sampled voxel signals enter convolution and region suggestions to obtain a three-dimensional bounding box.
The invention considers the relation expression between the original points of the enhanced bottom layer in the characteristic extraction process, utilizes the thought of a graph network in the characteristic extraction process, and simultaneously considers an attention mechanism imitating human cognitive acuity for better enhancing the characteristic expression, thereby leading the multi-channel selection of the characteristics to be more intelligent. The invention applies the attention mechanism before the initial feature selection of the graph network edge convolution and after the sparse convolution feature graph processing, and ensures that the feature expression of each stage is more explanatory while improving the expressive force of the neural network module.
The invention has the following advantages:
1. the invention uses the graph rolling method of the attention mechanism in the characteristic representing process of each voxel, can better describe the relation between each point of the point cloud and extract the characteristic with more expressive force.
2. According to the invention, after the intermediate layer convolution, the weight of the obtained high-order feature map is redistributed by using a attention mechanism, so that a more reasonable high-order feature map is obtained.
3. The two improvements work together, and the method can improve the accuracy of three-dimensional target detection in the detection of the vehicle.
Drawings
Fig. 1: a feature extraction module based on a graph network attention structure, wherein e represents an edge, x represents a point, and i and j represent the numbers of the points;
fig. 2: extracting voxel characteristics;
fig. 3: middle layer sparse convolution introducing attention mechanism;
fig. 4: overall flow.
Detailed Description
The invention is described in more detail below by way of example.
Step one: voxel division clustering of point cloud
And structuring and downsampling point cloud data of more than 100k points in a voxelized mode, firstly cutting out points out of a certain range, and only reserving points in D, H and W in x, y and z axes. Because the number of points of a pair of point clouds is too large, the size v is utilized within the extraction range d ，v h ，v w The whole point cloud is divided by the small voxel grid of (a).
In order to solve the problem of uneven distribution of points in each voxel, the present embodiment uses a method of random downsampling so that the number of points in each voxel does not exceed T. And finally numbering the processed voxel structure, and storing the voxel structure in a hash table mode, so that voxels with empty internal points are eliminated.
Step two: point cloud feature extraction in voxels
After voxelization of the original point cloud, the present embodiment performs feature extraction for each voxel using a graph annotation network module in order to obtain voxel-level features.
The point cloud is a natural graph structure, and in the feature extraction of the point cloud, each point is conventionally considered independently to neglect the relation between the points, so as to define
Is a graph comprising a set of n points
And edge set between points ++>
For example, the present invention defines a d-dimensional proximity graph for each point x i In->
Comprises (i, j) i1 )，...，(i，j ik ) An edge set of the form where i and j are both the numbers of points, thus defining an edge feature +.>
Wherein h is θ And H in the following formula is a symmetric function.
In general, a point cloud has three dimensions to represent its real world coordinates, and in this embodiment, the center point x is combined when describing the edge between two points i And a point connected thereto by h operation
As an initial feature selection. At this point, each channel of the edge feature contributes differently to the overall feature representation, and thus an attention mechanism approach is added. After the multi-layer sensing operation of the edge convolution, the edge-level features are extracted by using a symmetrical operation H, and the corresponding point-level features are obtained. Subsequently, by combining the point-level feature x= { X' 1 ，...，x′ n And performing another symmetrical operation to extract the final voxel level characteristics.
Step three: middle layer sparse convolution
The present embodiment uses a three-dimensional sparse convolution operation as the convolution intermediate layer. Suppose ConvMD (c) in ，c out K, s, p) is a convolution operator, where c in And c out Is the number of inputs and the output channels, k, s, p, corresponds to the kernel size, stride size, and fill size, respectively. Each convolution operation contains a 3D convolution, a batch normal layer and a Relu layer. Finally, after converting the sparse map to the dense map, an advanced feature map is obtained, and an attention module is added thereto.
There are many different scale patterns during the convolution operation. It is apparent that the contribution of the features of each dimension to the overall feature has a different importance. To improve the description of the feature map, and make it more reasonable, the present invention adds attention to the feature map to the original feature map.
The present embodiment uses an SE attention module for generating an attention profile. First, let dense feature map input be
Where H is the feature map height, W is the feature map width, and C is the channel number. Each channel is then extracted using an avg-pool operation to obtain an extracted feature, thus obtaining statistically derived channel weights
Multilayer perceptions are then used to obtain some advanced features for each dimension, with the final attention being given to s c ＝F e (z c W), where F e To extract a function.
At the scaling function F scale After that, attention feature map is added to the original map to obtainFinal output comprehensive characteristic diagram
This attention mechanism operation is added after the middle layer, and advanced information can be aggregated into the final middle layer feature map to provide more information for subsequent regional suggestions.
Step four: regional advice network
Regional recommendation networks (RPNs) have become a typical embedded module in many detection frameworks. In this embodiment, an end-to-end form resembling SSD is used as the region suggestion architecture. The input of the region suggestion layer is a feature map extracted by the middle layer, and one region suggestion layer contains a convolution layer, a Batchnormal layer and a Relu layer. After each individual RPN layer, the feature maps are up-sampled to the same fixed size and the maps are concatenated together. Finally, three 1×1 convolutions are used to generate the predicted values for bounding boxes, classes and directions.
Claims (1)
1. A three-dimensional target detection method based on a graph convolution attention network is characterized by comprising the following steps of:
step one: voxel division clustering of point cloud
Structuring and downsampling original point cloud data in a voxelized mode, discarding outliers outside a specified range, dividing the point cloud into grids, randomly downsampling in each voxel grid, numbering each grid, and storing; using a method of random downsampling, so that the number of points in each voxel is not more than T; finally numbering the processed voxel structure, and storing the voxel structure in a hash table mode, so that voxels with empty internal points are eliminated;
step two: point cloud feature extraction in voxels
After voxelization of the original point cloud, extracting features of each voxel by using a graph annotation network module in order to obtain voxel level features;
the point cloud is a natural graph structure, and each point is conventionally considered independently and negligibly in the feature extraction of the point cloudDefinition of Point-to-Point relationship
Is a graph comprising a set of n points
And edge set between points ++>
The point cloud has three dimensions to represent its real world coordinates, and after the multi-layer perceptive operation of edge convolution, a symmetry operation H is used to extract edge-level features, which are obtained by extracting the corresponding point-level features by using the point-level features x= { X' 1 ，...，x′ n Performing another symmetrical operation extraction to obtain final voxel level characteristics;
step three: middle layer sparse convolution
Using three-dimensional sparse convolution operation as a convolution intermediate layer, assume ConvMD (c in ，c out K, s, p) is a convolution operator, where c in And c out The number of inputs and the output channels, k, s, p correspond to kernel size, stride size, and fill size, respectively; each convolution operation includes a 3D convolution, a batch normal layer and a Relu layer; after the sparse mapping is converted into dense mapping, an advanced feature mapping is obtained, and an attention module is added;
using SE attention module for generating attention feature map, first, let dense feature map input as
Wherein H is the height of the feature map, W is the width of the feature map, and C is the number of channels; each channel is then extracted using an avg-mapping operation to obtain an extracted feature, thus obtaining statistically derived channel weights
Multilayer perceptions are then used to obtain some advanced features for each dimension, with the final attention being given to s c ＝F e (z c W), where F e Is an extraction function;
at the scaling function F scale Thereafter, the attention feature map is added to the original map to obtain a final output integrated feature map
Step four: regional advice network
Using an end-to-end form similar to SSD as a region suggestion architecture, wherein the input of a region suggestion layer is a feature map extracted by an intermediate layer, and one region suggestion layer comprises a convolution layer, a Batchnormal layer and a Relu layer; after each individual RPN layer, upsampling the feature maps to the same fixed size and concatenating the maps together; finally, three 1×1 convolutions are used to generate the predicted values for bounding boxes, classes and directions.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910918980.6A CN110674829B (en) | 2019-09-26 | 2019-09-26 | Three-dimensional target detection method based on graph convolution attention network |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910918980.6A CN110674829B (en) | 2019-09-26 | 2019-09-26 | Three-dimensional target detection method based on graph convolution attention network |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN110674829A CN110674829A (en) | 2020-01-10 |
| CN110674829B true CN110674829B (en) | 2023-06-02 |
Family
ID=69079355
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201910918980.6A Active CN110674829B (en) | 2019-09-26 | 2019-09-26 | Three-dimensional target detection method based on graph convolution attention network |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN110674829B (en) |
Families Citing this family (18)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111340189B (en) * | 2020-02-21 | 2023-11-24 | 之江实验室 | Space pyramid graph convolution network implementation method |
| CN111401190A (en) * | 2020-03-10 | 2020-07-10 | 上海眼控科技股份有限公司 | Vehicle detection method, device, computer equipment and storage medium |
| CN111583263B (en) * | 2020-04-30 | 2022-09-23 | 北京工业大学 | Point cloud segmentation method based on joint dynamic graph convolution |
| CN111476843B (en) * | 2020-05-08 | 2023-03-24 | 中国科学院合肥物质科学研究院 | Chinese wolfberry branch recognition and positioning method based on attention mechanism and improved PV-RCNN network |
| CN111539949B (en) * | 2020-05-12 | 2022-05-13 | 河北工业大学 | Point cloud data-based lithium battery pole piece surface defect detection method |
| WO2021226876A1 (en) * | 2020-05-13 | 2021-11-18 | 华为技术有限公司 | Target detection method and apparatus |
| CN116132671A (en) * | 2020-06-05 | 2023-05-16 | Oppo广东移动通信有限公司 | Point cloud compression method, encoder, decoder and storage medium |
| CN113971712A (en) * | 2020-07-22 | 2022-01-25 | 上海商汤临港智能科技有限公司 | Point cloud data processing method and device, electronic equipment and storage medium |
| CN112270289A (en) * | 2020-07-31 | 2021-01-26 | 广西科学院 | Intelligent monitoring method based on graph convolution attention network |
| CN112184867B (en) * | 2020-09-23 | 2024-09-17 | 中国第一汽车股份有限公司 | Point cloud feature extraction method, device, equipment and storage medium |
| CN112115954B (en) * | 2020-09-30 | 2022-03-29 | 广州云从人工智能技术有限公司 | Feature extraction method and device, machine readable medium and equipment |
| CN112257852B (en) * | 2020-11-04 | 2023-05-19 | 清华大学深圳国际研究生院 | Method for classifying and dividing point cloud |
| CN112633376A (en) * | 2020-12-24 | 2021-04-09 | 南京信息工程大学 | Point cloud data ground feature classification method and system based on deep learning and storage medium |
| CN112446385B (en) * | 2021-01-29 | 2021-04-30 | 清华大学 | Scene semantic segmentation method and device and electronic equipment |
| CN112862719B (en) * | 2021-02-23 | 2022-02-22 | 清华大学 | Laser radar point cloud cell feature enhancement method based on graph convolution |
| CN113900119B (en) * | 2021-09-29 | 2024-01-30 | 苏州浪潮智能科技有限公司 | Method, system, storage medium and equipment for laser radar vehicle detection |
| CN114266992B (en) * | 2021-12-13 | 2024-10-15 | 北京超星未来科技有限公司 | Target detection method and device and electronic equipment |
| CN115273645B (en) * | 2022-08-09 | 2024-04-09 | 南京大学 | Map making method for automatically clustering indoor surface elements |
Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN104573731A (en) * | 2015-02-06 | 2015-04-29 | 厦门大学 | Rapid target detection method based on convolutional neural network |
| CN109685813A (en) * | 2018-12-27 | 2019-04-26 | 江西理工大学 | A kind of U-shaped Segmentation Method of Retinal Blood Vessels of adaptive scale information |
Family Cites Families (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108520275A (en) * | 2017-06-28 | 2018-09-11 | 浙江大学 | A kind of regular system of link information based on adjacency matrix, figure Feature Extraction System, figure categorizing system and method |
| US11556777B2 (en) * | 2017-11-15 | 2023-01-17 | Uatc, Llc | Continuous convolution and fusion in neural networks |
| CN109255791A (en) * | 2018-07-19 | 2019-01-22 | 杭州电子科技大学 | A kind of shape collaboration dividing method based on figure convolutional neural networks |
| CN109934826B (en) * | 2019-02-28 | 2023-05-12 | 东南大学 | Image feature segmentation method based on graph convolution network |
| CN110222653B (en) * | 2019-06-11 | 2020-06-16 | 中国矿业大学(北京) | Skeleton data behavior identification method based on graph convolution neural network |
-
2019
- 2019-09-26 CN CN201910918980.6A patent/CN110674829B/en active Active
Patent Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN104573731A (en) * | 2015-02-06 | 2015-04-29 | 厦门大学 | Rapid target detection method based on convolutional neural network |
| CN109685813A (en) * | 2018-12-27 | 2019-04-26 | 江西理工大学 | A kind of U-shaped Segmentation Method of Retinal Blood Vessels of adaptive scale information |
Non-Patent Citations (2)
| Title |
|---|
| Can Chen等.GAPNet: Graph Attention based Point Neural Network for Exploiting Local Feature of Point Cloud.《arXiv》.2019,第1-11页. * |
| Zongji Wang等.VoxSegNet: Volumetric CNNs for Semantic Part Segmentation of 3D Shapes.《IEEE Transactions on Visualization and Computer Graphics》.2019,第26卷(第9期),第2919 - 2930页. * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN110674829A (en) | 2020-01-10 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN110674829B (en) | Three-dimensional target detection method based on graph convolution attention network | |
| CN109685842B (en) | Sparse depth densification method based on multi-scale network | |
| CN110443842B (en) | Depth map prediction method based on visual angle fusion | |
| CN111160214B (en) | 3D target detection method based on data fusion | |
| Guerry et al. | Snapnet-r: Consistent 3d multi-view semantic labeling for robotics | |
| CN110992271B (en) | Image processing method, path planning method, device, equipment and storage medium | |
| CN109583483B (en) | Target detection method and system based on convolutional neural network | |
| CN110659664B (en) | SSD-based high-precision small object identification method | |
| CN114708585A (en) | Three-dimensional target detection method based on attention mechanism and integrating millimeter wave radar with vision | |
| CN113160062B (en) | Infrared image target detection method, device, equipment and storage medium | |
| CN111046767B (en) | 3D target detection method based on monocular image | |
| CN114758337A (en) | Semantic instance reconstruction method, device, equipment and medium | |
| US20230153965A1 (en) | Image processing method and related device | |
| CN110956119B (en) | Method for detecting target in image | |
| CN114764856A (en) | Image semantic segmentation method and image semantic segmentation device | |
| CN113592015B (en) | Method and device for positioning and training feature matching network | |
| CN117422884A (en) | Three-dimensional target detection method, system, electronic equipment and storage medium | |
| CN112991537A (en) | City scene reconstruction method and device, computer equipment and storage medium | |
| CN116152800A (en) | 3D dynamic multi-target detection method, system and storage medium based on cross-view feature fusion | |
| CN115457539A (en) | 3D target detection algorithm based on multiple sensors | |
| CN114638866A (en) | Point cloud registration method and system based on local feature learning | |
| CN113569896A (en) | Computer-implemented method for 3D positioning of objects based on image and depth data | |
| CN111435448A (en) | Image salient object detection method, device, equipment and medium | |
| CN114219757B (en) | Intelligent damage assessment method for vehicle based on improved Mask R-CNN | |
| CN116957999A (en) | Depth map optimization 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 |