Visual Navigation Based On Floor Feature Segmentation For Indoor Mobile Robot

Vision has the advantage of broad sensing area and full information. With the development of image processing technology and the ability of computer, the visual navigation is becoming a main way in robot navigation. The progress made in the last two decades has been on vision-based navigation both for indoor robots and for outdoor robots. One task of the robot navigation is avoiding obstacles. Hence, robots should recognize the drivable and the obstacle areas as they move.With the support of project of National Scientific Fund and project of Excellent Scientist Fund in Hubei, the research on floor segmentation for indoor mobile robot visual navigation is put forward. This paper studies image processing theory, especially the color image segmentation. It puts forward a method of segmenting the scene into drivable and non-drivable areas through analyzing the color information of the input environment images and implements a system of obstacle detection, including the capture, processing, compression, saving and rendering of video images.Through the study of this subject, some results and conclusions are drawn:(1) This paper analyzes the features of corners, borders and color of indoor floor in detail. It also compares the methods of extraction these features. Finally the color feature is chosen as the landmark for visual navigation.(2) This paper does research on color image segmentation based on existing image segmentation algorithms. It puts forward a color image segmentation algorithm based on histogram to segment the indoor scene into drivable and non-drivable areas.(3) This paper implements a real-time system of video capture, processing, compression, saving and rendering based on DirectShow. The processed video images can be compressed by many methods.Visual C++ environment, OpenCV and DirectShow tools are used. The results show that this system can implement real-time obstacle detection and avoidance.