With the wide application of intelligent manufacturing system,the dem and of industrial robots is increasing rapidly.The existing robot parts product ion line can not judge whether the parts are qualified or not,which leads to th e complicated test process and low qualification rate of robot parts.In order t o solve the above problems,this paper proposes a recognition method of robot wr ist pinions based on fusion features,which can effectively identify and judge w hether the contour and texture of the wrist pinions of the robot are qualified o r not.Firstly,the full view image sample database of the target is established by using the simulation software of 3ds max,and then the Zernike moment featur e extraction and hog feature extraction are carried out for the wrist pinions of industrial robot.Thirdly,the dimension of the extracted high-dimensional fea t ures are reduced by different methods.Finally,the fusion features are trained and recognized by BP neural network classifier.The experimental results show th at:the recognition characteristics of robot wrist pinions based on fusion featu res are better than the traditional single feature recognition,and achieve coor dination in accuracy,real-time,stability,robustness and other aspects,and t h e comprehensive recognition effect is good,which is suitable for the qualified judgment of robot parts contour and texture.