Assembly Defect Detection of Atomizers Based on Machine Vision

Abstract

Atomizers are assembled in an automated assembly line, which inevitably creates assembly defects. In this paper, we use machine vision technology to detect assembly defects in atomizers. We propose two algorithms: an image processing algorithm, and a deep learning algorithm based on convolutional neural network. For design of the image processing algorithm, we set the region of interest for detection according to the position of different assembly defects. For the deep learning algorithm, we adopt the MobileNet model and propose a new training program to improve detection accuracy. The paper also includes an evaluation of the performance of the two algorithms and analyzes their advantages and disadvantages.