Object Detection with optimized tracker is a complex process because it faces many problems such as occlusion, blur image, background clutter, fast motion and many obstacles.  Tracker which is based on Reinforcement learning can solve this problem, RODRLM (Real Time Object Detection Deep Reinforcement Learning Model) based on reinforcement learning can track the object with great improvement of previous tracker. Although ADNet has some limitation in optimal action selection, and suffer from inefficient tracking .So some improvement is supposed to improve ADNet to enhance the tracking accuracy and efficiency. Firstly, the multi-domain training is incorporated into ADNet to further improve the feature extraction ability of its convolution layers. Then, in the reinforcement learning based training phase, both the selection criteria for optimal action and the reward function are redesigned separately to explore more appropriate action and eliminate useless action. an effective online adaptive update strategy is proposed to adapt to the appearance changes or deformation of the object during actual tracking.

Index Terms- Computer Vision, Deep Learning, Reinforcement Learning, Visual Tracking.