Research Of The Object Search And Recognition For A Mobile Robot Based On A Hybrid Vision System

The development of science and technology and the trend of aging of population have brought wide application for service robotics. As the vision based object search and recognition problem is now hotspot in the study of high level control of mobile robots. Therefore, it makes great sense to realize the object search and recognition effectively.In recent literatures, many robots utilize conventional visual system to get the environment information, due to the limited field of view, the conventional visual system can only obtain local scene information and require extensive image processing, with reduce the efficiency of target search. So, combined with omni-directional vision system and Kinect sensor, omni-directional vision as the main source of scene information, the corresponding key technologies are analyzed, such as: the object localization, search path planning and object recognition. The major works are listed as follows:(1)The Optical imaging model and epipolar geometry mode of the hyperbolic omni-directional vision and Kinect sensor is introduced, and the vision sensor is calibrated and some experiments of the spatial point localization based on an omni-directional vision are finished.(2) To the problem that the color characteristics of the area and color are sensitive to the bad lighting conditions, a search method based on the HSV color space algorithm is proposed.(3) According to the detection range of Kinect sensor and a positive observation object, determination of local robot detect pose, so as to complete the robot’s position adjustment to the next observation point.(4) According to the target for the positive images, combined with the Scale-Invariant Feature Transform and posture method, to identify the target object by matching between with the target and template.(5) Moving cost is the criteria to determine the detection order and colony algorithm is used to find the shortest path.