Paper Info
Title
Improved Multi-Object Tracking Algorithm For Forward Looking Sonar Based On Rotation Estimation
Abstract
Multi-object tracking algorithm for forward-looking sonar (FLS) often requires carrier motion information, in order to correct the tracking error caused by the carrier motion. However, it is sometimes difficult to obtain carrier information or synchronize the sonar image with carrier attitude sensor data. Therefore, it is still meaningful to study tracking multiple objects without using carrier motion information. In this paper we focus on improving the performance of multi object tracking without navigation data when the sonar carrier is rotating. The traditional detection-by-tracking framework was improved by rotation estimation. A linear motion model in polar coordinate system was used, and both detection and data association were performed in polar coordinates system. Phase correction was used to estimate the rotation velocity. The velocity was added to the motion model as a control. The improved method was tested on real sonar sequence obtained in conditions which carrier rotated several times. For evaluating the algorithm, we presented a simple approach for object detection. Finally, the results of several tracking metrics show better performance compared with conventional tracking method.
Year
DOI
Venue
2019
10.1007/978-3-030-27532-7_15
INTELLIGENT ROBOTICS AND APPLICATIONS, ICIRA 2019, PT II
Keywords
DocType
Volume
Forward looking sonar, Multi-object tracking, Motion estimation
Conference
11741
ISSN
Citations 
PageRank 
0302-9743
0
0.34
References 
Authors
0
2
Name
Order
Citations
PageRank
Xiufen Ye14210.31
Xinglong Ma200.34