Research Of Object Perception Based On Vision For Home Service Robot

With the development of society and the aging problem becoming more severe,the demand for home service robots is increasingly urgent.Object perception is the foundation for home service robots achieving intelligent job,as well as the hotspots in machine vision.Several factors,such as the unstructured indoor environment,various categories of objects,arbitrarily placed poses,contacts and occlusions between objects,image blurring caused by robot motion and the real-time requirement,bring a lot of difficulties for object perception.This thesis focuses on the challenging problem of service robots.The main research contents are as follows:Firstly,the object recognition system for home service robots based on SURF was designed and implemented.The system mainly includs three processes: feature point detection,feature point description and feature points matching.To solve the problem of false matching,RANSAC algorithm was used for matching purification,which can effectively eliminate the mismatch.The hardware platform of the robot was built.Experiments verify the robustness of this method,but the real-time performance needs to be improved.Secondly,in order to further enhance the performance,an object recognition algorithm for robots based on saliency region guidance was proposed.Salient region was extracted from the image based on visual selective attention mechanism.The SURF feature was extracted from the salient region.By matching the feature points,the recognition of target object in the scene was realized.Experimental results show that the recognition rate and the speed are effectively improved compared with the traditional SURF algorithm.Finally,in order to solve the problem that the traditional Camshift algorithm needs to manually select the tracking target,an unsupervised object tracking method for home service robots based on detection was proposed,which can search the target object automatically by SURF feature matching.So the automatic tracking was realized.To tackle the problem that the traditional Camshift algorithm is easy to be interfered by the similar color in the background,the SURF feature was integrated into the Camshift algorithm frame,and a new object tracking method combined with SURF feature was proposed.Experimental results show that the algorithm can effectively solve the problem of tracking failure in the background similar to object color of traditional Camshift algorithm.