Aimed at the problems of slow running speed and poor matching accuracy of traditional workpiece image recognition algorithms, an improved ORB (Oriented FAST and Rotated BRIEF) algorithm was proposed to solve the realtime and accurate recognition of workpiece images. The process of the algorithm was that the ORB algorithm was used to extract the corner point features of the workpiece image,then SURF (speed-up Robust Features) descriptors were added for direction allocation to obtain image corner points with rotation scale invariant, and the feature points were matched with the fast approximate nearest neighbor search algorithm to realize the identification of the workpiece image.The experimental results show that the improved ORB algorithm is faster and more accurate than the traditional ORB, SIFT (Scale Invariant Feature Transform) and SURF algorithms, SIFT+SURF and SURF+FREAK combination algorithm in the workpiece image corner points extraction and target matching in the case of rotation scale variation of the image. Image recognition efficiency and accuracy of industrial robot in the process of moving parts are improved