High-precision and real-time visual tracking algorithm based on the Siamese network for autonomous driving

Visual object tracking is often used to track obstacles in autonomous driving tasks. It requires real-time performance while dealing with target deformation and illumination changes. To solve the above problems, this paper proposes a high-precision and real-time visual tracking algorithm for autonom...

Full description

Saved in:
Bibliographic Details
Published in:Signal, image and video processing image and video processing, 2023-06, Vol.17 (4), p.1235-1243
Main Authors: Lyu, Pengfei, Wei, Minxiang, Wu, Yuwei
Format: Article
Language:English
Subjects:
Algorithms
Artificial neural networks
Computer Imaging
Computer Science
Datasets
Feature extraction
Image Processing and Computer Vision
Multimedia Information Systems
Optical tracking
Original Paper
Pattern Recognition and Graphics
Qualitative analysis
Real time
Signal,Image and Speech Processing
Vision
Citations: Items that this one cites
Items that cite this one
Online Access:Get full text
Tags: Add Tag
No Tags, Be the first to tag this record!
Description
Summary:Visual object tracking is often used to track obstacles in autonomous driving tasks. It requires real-time performance while dealing with target deformation and illumination changes. To solve the above problems, this paper proposes a high-precision and real-time visual tracking algorithm for autonomous driving based on the Siamese network. First, our tracker utilizes ensemble learning to fuse two feature extraction branches that are derived from the convolutional neural network. Then, the channel attention mechanism is added before concatenation to redistribute feature weights. Finally, a region proposal network is adopted to generate tracking bounding boxes. Extensive experiments demonstrate that compared with the state-of-the-art algorithms, the proposed method achieves satisfactory results on four benchmark datasets while maintaining a higher frame rate. Also, the qualitative analysis results on the KITTI dataset indicate that our method can meet the challenges in autonomous driving.
ISSN:1863-1703
1863-1711
DOI:10.1007/s11760-022-02331-y