Evaluation of tracking performance: focused on improvement of aiming ability for individual weapon

Abstract

In the present study, we experimentally investigate to improve individual weapon aiming by using Visual Tracking. This was achieved through actual battlefield combat for a few days over a short period of time. The forces were involved in situations where continued operation is reduced by various elements. In the future, operations in the battlefield will be even more critical due to nuclear, biological, and chemical weapons. We improved the aiming through roof mounted sight with visual tracking technology that automatically tracked troops. We chose the incremental visual tracker, multiple instance learning, and visual tracking decomposition, which were published in IJCV and CVPR, as algorithms to apply in various battlefield sequences. These battlefield sequences include occlusion, moving cameras, scale change, small objects, low-contrast, and change in illumination. The experimental results showed that visual tracking decomposition was slow, but had the best performance.