A Target Positioning Method for Industrial Robot Based on Multiple Visual Sensors

Abstract

Flexible start-up is the future trend of production automation. To realize intelligent and flexible operations, industrial robot must position the target rapidly and accurately. Otherwise, it is impossible for the robot to operate automatically in complex environment. This paper designs a novel target positioning method that enables industrial robot to position its target with high precision and fast speed. Firstly, the target images were preprocessed through enhancement, histogram equalization, and filtering. Next, the target motion areas (TMAs) captured by the system of multiple visual sensors (MVSs) were subject to information fusion, and the feature points of fused image were matched and optimized. After that, the fused image was recognized and described by speeded up robust features (SURF)fast retina key-point (FREAK) algorithm. Finally, a two-dimensional (2D) data model was established based on the centroid coordinates of the fused image. Experimental results prove that our method can effectively and accurately position target in complex environment, while simplifying size measurement and speeding up computation. The research results provide a reference for image collection and information fusion by MVSs system in other fields.