US 12,437,366 B2
Target detection method, computer device, computer-readable storage medium, and vehicle
Guanghui Ren, Shanghai (CN); Huan Qin, Shanghai (CN); Xindong He, Shanghai (CN); Qi Xiong, Shanghai (CN); and Chao Peng, Shanghai (CN)
Assigned to Anhui NIO Autonomous Driving Technology Co., Ltd., Hefei (CN)
Filed by Anhui NIO Autonomous Driving Technology Co., Ltd., Hefei (CN)
Filed on May 22, 2023, as Appl. No. 18/321,438.
Claims priority of application No. 202210564922.X (CN), filed on May 23, 2022.
Prior Publication US 2023/0377105 A1, Nov. 23, 2023
Int. Cl. G06T 5/30 (2006.01); G06F 18/20 (2023.01)
CPC G06T 5/30 (2013.01) [G06F 18/29 (2023.01); G06T 2207/10028 (2013.01); G06T 2207/20216 (2013.01); G06T 2207/20221 (2013.01); G06T 2207/30252 (2013.01)] 15 Claims
OG exemplary drawing
 
1. A target detection method, comprising:
rasterizing point cloud space of three-dimensional point clouds in a vehicle driving environment to form a plurality of three-dimensional point cloud grids, and using point cloud grids comprising three-dimensional point clouds as target point cloud grids;
determining a convolution dilation rate corresponding to each of the target point cloud grids based on sparsity of a respective target point cloud grid, wherein the convolution dilation rate is positively correlated with the sparsity;
dilating a sparse convolution based on the convolution dilation rate to form a dilated sparse convolution;
extracting a point cloud grid feature of the corresponding target point cloud grid by using the dilated sparse convolution;
weighting the point cloud grid feature by using an attention mechanism to obtain a global point cloud feature; and
performing target detection based on the global point cloud feature.