Abstract:Object tracking as a low-level vision task has always been a hot topic in computer vision. It is well known that Challenges such as background clutters, fast object motion and occlusion et al. affect a lot the robustness or accuracy of existing object tracking methods. This paper proposes a reinforcement learning model based on Twin Delayed Deep Deterministic algorithm (TD3) for single object tracking. The model is based on the deep reinforcement learning model, Actor-Critic (AC), in which the Actor network predicts a continuous action that moves the target bounding box in the previous frame to the object position in the current frame and adapts to the object size. The Critic network evaluates the confidence of the new bounding box online to determine whether the Critic model needs to be updated or re-initialized. In further, in our model we use TD3 algorithm to further optimize the AC model by using two Critic networks to jointly predict the bounding box confidence, and to obtain the smaller predicted value as the label to update the network parameters, thereby rendering the Critic network to avoid excessive estimation bias, accelerate the convergence of the loss function, and obtain more accurate prediction values. Also, a small amount of random noise with upper and lower bounds are added to the action in the Actor model, and the search area is reasonably expanded in offline learning to improve the robustness of the tracking method under strong background interference and fast object motion. The Critic model can also guide the Actor model to select the best action and continuously update the state of the tracking object. Comprehensive experimental results on the OTB-2013 and OTB-2015 benchmarks demonstrate that our tracker performs best in precision, robustness, and efficiency when compared with state-of-the-art methods.

End-to-end active object tracking and its real-world deployment via reinforcement learning

End-to-End Active Object Tracking and Its Real-World Deployment Via Reinforcement Learning

End-to-end Active Object Tracking Via Reinforcement Learning

End-to-end Active Object Tracking Football Game Via Reinforcement Learning

Robust Object Tracking with a Hierarchical Ensemble Framework

Beyond Traditional Driving Scenes: A Robotic-Centric Paradigm for 2D+3D Human Tracking Using Siamese Transformer Network

Space Noncooperative Object Active Tracking with Deep Reinforcement Learning.

Space Non-cooperative Object Active Tracking with Deep Reinforcement Learning

End-to-End Binocular Active Visual Tracking Based on Reinforcement Learning

Anti-distractor Active Object Tracking in 3D Environments

Real-Time Visual Object Tracking Based On Reinforcement Learning With Twin Delayed Deep Deterministic Algorithm

Enhancing Continuous Control of Mobile Robots for End-to-End Visual Active Tracking

Deep Reinforcement Learning-Based DQN Agent Algorithm for Visual Object Tracking in a Virtual Environmental Simulation

Visual Tracking Using Online Deep Reinforcement Learning with Heatmap

Reinforcement-Tracking: An End-to-End Trajectory Tracking Method Based on Self-Attention Mechanism

Deep Active Contours for Real-time 6-Dof Object Tracking

On Deep Recurrent Reinforcement Learning for Active Visual Tracking of Space Noncooperative Objects

TrackAgent: 6D Object Tracking via Reinforcement Learning

Collaborative Deep Reinforcement Learning for Multi-object Tracking

Trustworthy Dynamic Object Tracking Using Deep Reinforcement Learning with the Self-Attention Mechanism

Hierarchical Active Tracking Control for UAVs via Deep Reinforcement Learning