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

This study mainly researches on the system of object recognition and pickup for mobile robot which is based on monocular vision. The system is in relation to color image segmentation, object recognition, camera calibration off-line, object location and distance measurement based on monocular and so on. By choosing appropriate algorithms and their combination, the system has achieved good result.Firstly, in the image segmentation module, a new segmentation method based on HSL color model is brought forward which improves former segment result. This method can distinguish different colors well, and is robust to different illumination conditions. If the object is placed in complex circumstance, by using the floodfill algorithm, most interferefering regions can be removed.Secondly, in the object recognition module, outer contour extraction algorithm is used on the segmented image, then contour-based moment invariants are used to match the target area with the template. This method reduces operating time greatly comparing to traditional region-based moment invariants.Thirdly, in the location and distance measurement module, a method using ground restriction model to measure distance is proposed, this method is derived from solid geometry and photogrammetry theoretics.