基于深度学习的汽车智能表面自动化检测系统设计

Design of auto-test system for vehicle intelligent surface based on deep learning

在汽车智能表面自动化检测中,传统机器视觉方法因易受光源、环境光影响,导致物体缺陷特征参数提取困难,检测准确性不高,本文采用协作式机器人作为运动控制,用压力检测单元、电流采集单元和图像检测单元组成检测系统,建立基于深度学习的汽车智能表面缺陷检测方法。结果表明,缺陷识别和缺陷分类的准确率达到95%,比传统方法高10%以上。该系统提高了表面检测的自动化程度,减少了人力成本,降低了人为操作的失误率,实现了汽车智能表面的高精度、高鲁棒性自动化检测。

Traditional machine vision methods are susceptible to light source and ambient conditions when testing the defects in vehicle intelligent surface,resulting in feature extraction difficult,poor robustness and accuracy.In this paper,a collaborative robot is used as the motion control system,and the detection system is composed of a pressure detection unit,a current acquisition unit and a image detection unit.A method based on deep learning network of detecting the defects of intelligent surface is proposed.The accuracy of detecting and classification of defect has been reached 95%,improved 10%compared with the traditional methods.This system improved the automation level,the test accuracy and robustness greatly.