Font Size: a A A

Research On Visual Object Tracking And Its Implementation In Mobile Robot

Posted on:2017-10-01Degree:MasterType:Thesis
Country:ChinaCandidate:X HeFull Text:PDF
GTID:2348330503970071Subject:Control Engineering
Abstract/Summary:PDF Full Text Request
The visual object tracking technology is one of the critical topics in computer vision, which receives much attention from engineers and scientific researchers, possesses a wide practical application and research prospect. Object tracking is a difficult challenge in complicated environment, due to numerous factors such as partial occlusion, illumination variation, pose change, complex motion, and background clutter, developing a robust and real-time tracking is still a challenging problem.Traditionally, most tracking algorithm use the single feature to describe the object. To resolve the insufficiency of traditional tracking algorithm in complicated background, we propose an object tracking algorithm based on multi- feature fusion, which the proposed algorithm improve the object feature variety and the robustness of algorithm.We introduce a global constraint measure innovatively, and propose a adaptive searching mechanism to speed the algorithm. Then, we put forward an object tracking algorithm based on multi- feature fusion and visual saliency. First, adopting a visual saliency mechanism for manipulating color histogram data to get saliency feature, we use a hybrid strategy incorporated the local feature with saliency information to describe the image, and then build the object-foreground model and object-background model. Furthermore, the dynamic feature is extracted by the bidirectional optical flow and error metric and is fused with the static feature which is extracted by the adaptive searching mechanism. Finally, based on the data of matching and tracking procedure in the previous frame, we evaluate the object 's scale, rotation and center, and obtain the new target location in the current frame. Experiments show that our algorithm can handling the tracking in the complicated background, and adapt to strong illumination change, partial occasion, fast motion, pose change and so on. Compared with the related algorithms, the high accuracy and robustness of the proposed algorithm are confirmed, and effectiveness and real-time of the method is validated.Based on the hardware environment of robot IN-RE, we design a motion control strategy. Using our object tracking algorithm and motion control strategy, we implement a robot object tracking system. Experimental results show that robot IN-RE can achieve the task of object tracking in real world.
Keywords/Search Tags:Visual Saliency, BRISK Feature, Consensus-based Matching and Clustering, Multi-Feature Fusion
PDF Full Text Request
Related items