Design of multi-dimensional intelligent detection platform for automotive motor commutator based on deep learning

Abstract This paper proposes a multi-dimensional intelligent detection system based on deep learning to solve the challenges of shape and position errors and surface defects detection of precision components in industrial scenes. By integrating convolutional neural network (CNN) and machine vision technology, an algorithm framework is constructed, including image acquisition, sub-pixel positioning, adaptive filtering, and multi-angle template matching, to realize the omni-directional defect detection of the top, side, and bottom of the component. Experiments show that the detection efficiency of the system is improved by 70.6% (20 pieces/min) compared with the manual method, and the recognition rate is stable at about 98%. It supports parameter adaptive measurement of multi-dimensional components. This research provides an innovative solution for high-precision online detection of industrial components.