Research on Surface Target Detection Algorithm Based on 3D Lidar
Authors:
- 18 June 2021
- conference paper
- Published by Institute of Electrical and Electronics Engineers (IEEE)
- p. 489-494
Abstract
3D Lidar is the key perception module of Unmanned Surface Vehicle (USV). Targets in the background of the water are affected by refracted light. The visual sensor is difficult to detect in special scenes, which affects the autonomous navigation and obstacle avoidance function. This paper proposes a 3D lidar-based VoxelNet detection algorithm for water surface targets. The sparse point cloud data on the water surface is divided into voxel form, and the hash table is input for efficient query, and the feature tensor is extracted through the feature learning layer and input into the convolutional layer to obtain the global target Information to achieve high-precision target detection. Experimental results show that the surface target detection algorithm based on 3D lidar improves 13.6% compared with the visual solution, which provides a more effective technical means for the intelligent process of USV.Keywords
This publication has 8 references indexed in Scilit:
- Channel-Based Network for Fast Object Detection of 3D LiDARElectronics, 2020
- PointPillars: Fast Encoders for Object Detection From Point Clouds2019
- SECOND: Sparsely Embedded Convolutional DetectionSensors, 2018
- 3D Semantic Segmentation with Submanifold Sparse Convolutional Networks2018
- Multi-view 3D Object Detection Network for Autonomous Driving2017
- Vehicle Detection from 3D Lidar Using Fully Convolutional Network2016
- Unmanned surface vehicles: An overview of developments and challengesAnnual Reviews in Control, 2016
- VoxNet: A 3D Convolutional Neural Network for real-time object recognition2015