Implementation of Vision-Based Tracking System for a Quadrotor
This paper presents an object tracking platform based on computer vision. A combination of CamShift and Kalman filter algorithms is used to achieve real-time tracking of moving objects’ requirements. We proposed an effective control strategy to realize th
- PDF / 947,350 Bytes
- 9 Pages / 439.37 x 666.142 pts Page_size
- 116 Downloads / 229 Views
Abstract This paper presents an object tracking platform based on computer vision. A combination of CamShift and Kalman filter algorithms is used to achieve real-time tracking of moving objects’ requirements. We proposed an effective control strategy to realize the tracking algorithm. A tracking system consisting of a low-cost quadrotor and a control station is designed for validating the proposed approach. The scene of an AR.Drone quadrotor searching and following the target in a 3D space is presented at last. Keywords Object tracking
CamShift Kalman filter Quadrotor
1 Introduction As technologies of new material, micro-electromechanical systems (MEMS), and micro inertial measurement unit (MIMU) advance, unmanned aerial vehicle (UAV) has been developing rapidly and attracting lots of attention in both industrial and military fields. Quadrotor is considered to be the best platform in the field of UAV application and experiment. Because of the quadrotor has simple structure and good stability; many institutions use it to verify their flight control algorithm. For autonomous UAV, the ability of visual tracking is of vital importance. Many researches have designed various kinds of visual tracking system. Bi and Duan [1] present a tracking and landing approach mainly based on RGB color information; Chen et al. [2] develop an indoor trajectory-following algorithm and object-following method using monocular vision; Krajnik et al. [3] use an
X. Ding (&) Y. Yu College of Electronic and Information Engineering, Tongji University, Shanghai 201804, China e-mail: [email protected]
Z. Wen and T. Li (eds.), Foundations of Intelligent Systems, Advances in Intelligent Systems and Computing 277, DOI: 10.1007/978-3-642-54924-3_82, Springer-Verlag Berlin Heidelberg 2014
879
880
X. Ding and Y. Yu
AR.Drone quadrotor to perform basic tasks of position stabilization, object following, and autonomous navigation. In this work, the quadrotor is controlled to track a moving object in a 3D space. We present a new solution for autonomous tracking. The vision-based approach utilizes histogram rather than RGB values of target image to match the template [4]. A composite algorithm combining Kalman filter and CamShift is used to solve the tracking problem. Using the monocular camera onboard for image capturing and ground control station for image processing, the visual tracking system is implemented in a type of low-cost quadrotor AR.Drone.
2 The Solution for Autonomous Tracking 2.1 CamShift Algorithm CamShift algorithm is the extension of mean shift algorithm. It is based on the color distribution information to track the moving objects. The principle of CamShift is described as follows. According to the color histogram of target, it converts the image to color probability distribution and then applies mean shift algorithm to all the frames of video sequence. Through adaptively adjusting the size and position of search window, we can locate the center of the target in the current image [5]. Firstly, calculate the color histogram
Data Loading...
