Abstract
In remote sensing images, small target size and diverse background cause difficulty in locating targets accurately and quickly. To address the lack of accuracy and inefficient real-time performance of existing tracking algorithms, a multi-object tracking (MOT) algorithm for ships using deep learning was proposed in this study. The feature extraction capability of target detectors determines the performance of MOT algorithms. Therefore, you only look once (YOLO)-v3 model, which has better accuracy and speed than other algorithms, was selected as the target detection framework. The high similarity of ship targets will cause poor tracking results; therefore, we used the multiple granularity network (MGN) to extract richer target appearance information to improve the generalization ability of similar images. We compared the proposed algorithm with other state-of-the-art multi-object tracking algorithms. Results show that the tracking accuracy is improved by 2.23%, while the average running speed is close to 21 frames per second, meeting the needs of real-time tracking.
Highlights
With rapid developments in space science, remote sensing technology has greatly improved the small coverage of traditional ground detection and the lack of related data through the high-speed acquisition of omnidirectional and multi-view ground information.Ships are indispensable strategic resources and means of transportation in military and civilian fields
Remote sensing images of ships are small in size and have complex backgrounds, which makes ship target tracking more challenging than multi-object tracking of pedestrians on the road
Deep learning has greatly improved the performance of computer vision tasks due to its powerful feature extraction capabilities, but research related to multi-object tracking is minimal
Summary
With rapid developments in space science, remote sensing technology has greatly improved the small coverage of traditional ground detection and the lack of related data through the high-speed acquisition of omnidirectional and multi-view ground information. Deep learning has greatly improved the performance of computer vision tasks due to its powerful feature extraction capabilities, but research related to multi-object tracking is minimal. Seyed et al [5] first proposed an online multi-object tracking method based on a fully developed, end-to-end learning algorithm in 2016, which solved the problem of modeling target number changes and discrete data association. The DeepSORT algorithm is proposed based on the SORT algorithm, and multi-object tracking is improved significantly by extracting deep feature information. With rapid developments in deep learning, the convolutional neural network (CNN) algorithm has become the preferred framework for target detection networks due to its powerful feature extraction and modeling capabilities. YOLOv3 is significantly better than other algorithms in detecting accuracy and speed; a multi-object tracking algorithm based on an improved YOLOv3 network is proposed .
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have