Enhancing 3D Single Object Tracking with Efficient Point Cloud Segmentation

3D single object tracking (SOT) based on point cloud has attracted much attention due to its important role in machine vision and autonomous driving. Recently, M 2 -Track proposes a two-stage tracking structure centered on motion, but they ignore the effect of segmentation errors in sparse point cloud scenarios, which hinder the ability of networks to accurately represent tracking targets. To solve the problems, we propose an efficient 3D single object tracker (Abbr. EST) that can effectively segment point cloud features. Firstly, the proposed fusion segmentation module makes up for the feature loss caused by the downsampling strategy and enhances the ability of the network to recognize foreground points. In addition, the global embedded module is used to further focus on the crucial features of the target. This module provides global information by using residual networks and adding background information. Numerous experiments conducted on KITTI and NuScenes benchmarks show that EST achieves superior point cloud tracking in both performance and efficiency.

Authors

• Yushi Yang
• Baojie Fan
• Yuyu Jiang
• Wuyang Zhou
• Dong Chen
• Hongxin Xu

Keywords

• Point cloud compression
• Three-dimensional displays
• Target tracking
• Motion segmentation
• Object segmentation
• Benchmark testing
• Object tracking
• Optimization
• Intelligent robots
• Residual neural networks
• Point Cloud
• Single Tracking
• Point Cloud Segmentation
• Single Object Tracking
• Background Information
• Target Features
• Residual Network
• Ability Of The Network
• Segmentation Errors
• Machine Vision
• Sparse Point Cloud
• Segmentation Module
• Error Scenarios
• K-nearest Neighbor
• Pedestrian
• Local Point
• Attention Mechanism
• Intersection Over Union
• Multilayer Perceptron
• Bounding Box
• KITTI Dataset
• Average Precision
• Residual Block
• Point Cloud Data
• Changes In Appearance
• Regression Loss
• Subsequent Frames
• Aggregation Function
• Segmentation Results
• Tracking Performance