Multi-mode Input Siamese Network for Tracking Object in Nighttime
Dunyun He, Zhen Yang, Haiqiao Wen, Zhijian Yin · 2021 IEEE 21st International Conference on Communication Technology (ICCT) · 2021
In recent years, there have been many algorithms based on siamese network in the visual object tracking to solve the visual object rapidly move, background clutter, motion blur, deformation, and other complex problems. However, aiming at the problem of visual object tracking in nighttime environment, there is little research on this problem by scholars. Therefore, we propose a multi-mode input fusion module embedded in the siamese network for tracking object in nighttime environment. Since there is no common public dataset for nighttime vision target tracking, OTB-100 dataset is darkened to generate different degrees of simulated nighttime scenes. Experimental results on different degrees of OTB-100 simulated nighttime scenes show that our method can obtain better precision and success rate.