Research And Implementation Of Target Recognition And Face Recognition For Home Service Robots In Complex Environment

Vision system is the mainly channel for home service robots to acquire information, and the ability to recognize target or family members quickly and accurately in Complex Environment is the basement for home service robots to complete their tasks, therefore, it has extremely significance to research and realize the target recognition and authentication based on face recognition. According to many specific issues in the practical application of home service robots and RoboCup robot competitions such as reducing the influence of light conditions and camera angle brought to target recognition and face recognition, the following works had been done in this thesis:Firstly, we introduced and compared several classical algorithms of target recognition and face recognition to analyze their advantages and disadvantages, accordingly,proposed and implemented target recognition algorithms and face recognition algorithms on the basis of home service robots.Secondly, in order to take full use of object shape and texture information, ensure the target translation invariance, scale invariance and rotation without distortion, and reduce the influences of vary light conditions or different camera angle to the target recognition system,we optimized the shape invariant moments and formed feature vector , based on the multi-target recognition with color information,and compared the improved Euclidean distance of feature vectors,accordingly,we get an efficient, rapid and robust target recognition algorithm to meet the home service robot in real time target recognition accurately.Finally, faced the low recognition rate in the case of shortage of total number of samples in the common face recognition system,we picked many effective Haar features between human face and non face to face fast detection and location by Adaboost classification method, and geometry normalized and gray normalized the face region of image to avoid the influences of indoor environments with different light conditions on the accuracy of face recognition , and remove the background interference with maximum extent possible. Then, in order to represent the features of human faces effectively and recognize face with high efficiency in the case of face samples shortage, we constructed feature vectors used eigenfaces, then reduced dimension fast through PCA, and formed a basis only used the first few eigenvectors corresponding to the first few largest eigenvalues . the experiments used ORL face database proved that the algorithm can improve the matching speed of eigenfaces effectively, and the home service robot can perform fast face recognition accurately in the case of face samples shortage.