Three-dimensional target detection method and system based on substream sparse convolution

Abstract

The invention provides a three-dimensional target detection method and system based on substream sparse convolution. The method comprises the following steps: obtaining initial point cloud data of a target scene; based on the initial point cloud data and the optimized point cloud voxel convolutional neural network, obtaining target point cloud features, and obtaining the optimized point cloud voxel convolutional neural network by optimizing the point cloud voxel convolutional neural network through substream sparse convolution; and performing target detection on the target scene according to the target point cloud features. According to the embodiment of the invention, the point cloud voxel convolutional neural network extraction method based on substream sparse convolution is constructed,further feature extraction is carried out by using the sparsity of the point cloud, and deep extraction can be carried out on local information more efficiently so that the efficiency and precision of three-dimensional target detection are improved.

Description

technical field [0001] The invention relates to the technical fields of robots and computer vision, in particular to a three-dimensional object detection method and system based on subflow sparse convolution. Background technique [0002] In the field of autonomous driving and robotics, it is often necessary to directly acquire and process 3D data information. 3D data information has more depth information than ordinary 2D data information. These depth information can eliminate a large number of segmentation uncertainties in 2D images. And can provide important geometric information. [0003] Since the 1990s, the 3D shape feature extraction algorithm has experienced more than 20 years of development, gradually transitioning from manual feature extraction to deep learning. There are four main ways for deep learning to be used in 3D: (a) multi-view-based convolution, (b) voxel-based convolution, (c) structured convolution and (d) direct convolution of point clouds . Among t...

AI Technical Summary

Problems solved by technology

[0007] The present invention provides a three-dimensional object detection method and system based on subflow sparse convolution, which is used to solve the defects of low efficiency and large memory space of the PVCNN method in the prior art, and realize fast and efficient three-dimensional object detection

Images