Light-weight UAV object tracking network based on strategy gradient and attention mechanism

Xia Hua,Xinqing Wang,Ting Rui,Faming Shao,Dong Wang

DOI: https://doi.org/10.1016/j.knosys.2021.107071

2021-07-01

Abstract:<p>Most existing object tracking methods have poor adaptability to complex scenes, and cannot achieve a good balance between tracking accuracy and real-time performance. To solve the above problems, this paper proposes a lightweight UAV object real-time tracking algorithm based on strategy gradient and attention. Firstly, a lightweight E-Mobile Net is designed as the backbone network of feature extraction; secondly, a feature enhanced attention assistant module is designed to enhance the adaptability and discrimination ability of the model; with multi-layer feature fusion regional suggestion network, foreground background classification and boundary box regression response map are obtained by cross-correlation, and the tracking results are calculated. The strategy network based on strategy gradient is used to optimize the template update and re detection strategy, which improves the overall tracking accuracy and efficiency of the model. Simulation experiments on an embedded device and multiple standard data sets show that compared with the current mainstream algorithms, the tracking accuracy is significantly improved 20%<span class="math"><math>∼</math></span>30%, the algorithm robustness also has obvious advantages, and the tracking speed on an embedded device is 56 <em>fps</em> can meet the real-time requirements.</p>

computer science, artificial intelligence

What problem does this paper attempt to address?

The problem this paper attempts to address is the poor adaptability of existing target tracking methods in complex scenarios, and the inability to achieve a good balance between tracking accuracy and real-time performance. To overcome these issues, the paper proposes a lightweight UAV (Unmanned Aerial Vehicle) target real-time tracking algorithm based on policy gradient and attention mechanism.

Specifically, the paper mainly addresses the following issues:
1. **Adaptability in complex scenarios**: Existing target tracking methods perform poorly when facing challenges such as complex backgrounds, occlusion, and target scale changes.
2. **Balance between tracking accuracy and real-time performance**: Existing methods either sacrifice accuracy to improve real-time performance or sacrifice real-time performance to improve accuracy.
3. **Computational resource limitations**: The computational resources of UAV platforms are relatively limited, making it more difficult to implement real-time algorithms.

To solve the above problems, the paper proposes the following three improvements:
1. **Design of a new lightweight neural network E-MobileNet**: As the backbone network for feature extraction, this network has lower computational complexity while maintaining sufficient feature representation capability.
2. **Introduction of an enhanced feature attention auxiliary module**: This module enhances the model's adaptability and discrimination ability by obtaining fused multi-layer response maps through a multi-layer feature fusion mechanism.
3. **Design of a policy network based on policy gradient**: Optimizes template update and re-detection strategies, improving overall tracking accuracy and efficiency.

Through these improvements, experimental results on multiple standard datasets show that the algorithm significantly improves tracking accuracy on embedded devices by 20%~30%, has a clear advantage in robustness, and achieves a tracking speed of 56 frames per second, meeting real-time requirements.