Omni-vision Navigation System Based On Embedded Multi-objects Tracker

Robotic navigation is a key problem in the robotic system. An ideal solution to the navigation problem has been seen as"vision navigation"which has the advantages of large information, fast sampling and simple setting. In this paper, an embedded multi-objects tracker based on Omni-directional vision system was provided to implement recognition and tracking for the sequential beacons array landmarks. A navigation system using the embedded tracker is also provided to achieve the robotic automatic navigation in the indoor structured environment. The main research work in this paper is as follow:1. Model and calibrate the Omni-vision system which has been built by the fisheye lens. The Omni-vision based on fisheye lens, used for engineering applications, includes two coordinates: the lens coordinates and the image coordinates. The method for the calibration of the five main parameters of the model is also provided. The modeling and calibration of the Omni-vision system is the basic and prerequisite for vision navigation.2. Design a new method for multiple targets tracking based on Particle Filter. We employed Particle Filter as the probabilistic framework and designed a new algorithm, the Feature Matching Embedded Particle Filter (FMEPF), to integrate advantages of Particle Filter and feature matching, for tracking multiple targets accurately and robustly.3. In order to tracking multi-objects in real-time without delay or loss of the image data, the tracking algorithm was transplanted into DSP system using the alternate implementation technology and optimize, as an embedded multi-objects tracker. Compared the common CPU based on x86 structure, the embedded tracker has smaller volume, lower power and higher integration, which could efficiently offer robotic navigation and vehicle-mounted system.4. Based on embedded multi-objects tracker, a navigation system was built for robotic tracking navigation for indoor structured environment. The tracking navigation, as a rigid navigation, guides the robot using the sequential landmarks array in the environment. The navigation system localized the space coordinates of the robotic according to the landmarks tracking by embedded tracker, and set the guiding path for robotic tracking navigation.5. Designed a Teaching-Playback navigator for robotic navigation. Teaching/Playback navigation as a flexible navigation mode has more research value and practical value. First, the navigation path was teached and a Teaching-Playback Database (TPD) was built according the landmarks information tracked by embedded tracker. As playback, the navigator matched the landmark coordinates in the TPD, and set the guiding path according the position of the robot for robotic Teaching-Playback navigation.