Research On Object Recognition And Location For Mobile Robot System Based On Monocular Vision

Mobile robot monocular vision system is now a hotspot in the field of mobile robotresearch. Just like humans, vision is also the most important source for robot to know theworld around; it can supply the information of the environment, such as the objects’ distanceand direction. Comparing with the panoramic vision and binocular vision, monocular visionhas several advantages, such as simple structure, easy to calibration and flexible motion. It issignificant to research on mobile robot monocular vision system.This study mainly researches on object recognition and location for mobile robot systemwhich is based on monocular vision. The system is in relation to color image segmentation,object recognition and tracking, camera calibration, object location and distance measurementbased on monocular and so on.Firstly, in the image segmentation module, a new segmentation method is broughtforward which improves former segment result. This method can distinguish different colorswell, and is robust to different illumination conditions. If the object is placed in complexcircumstance, by using the floodfill algorithm, most interfering regions can be removed.Secondly, in the object recognition module, outer contour extraction algorithm is used onthe segmented image. This method reduces operating time greatly comparing to traditionalregion-based.Thirdly, in the monocular measurement module, a method using space restriction modelsto measure distance is proposed, this method is derived from solid geometry andphotogrammetry theoretics. Then the ground restriction model is applied to the robot systemto measure distance.