复杂电子装备具有结构组成复杂、总装工序多、精度要求高、机电液高度耦合、在线检测困难等问题,如果出现零部件装配错误将导致返工。为了实现装配过程质量在线检测,设计了基于机器学习的图像检测系统。首先,根据装备尺寸设计了可移动图像采集装置及相机布置方案,能够适用于不同尺寸装备的图像采集需求;然后,针对采集到的图像存在的畸变问题进行校正,并通过图像融合实现分段图像的拼接;最后,采用深度学习算法实现图像的识别,并采用知识图谱实现检测结果还原。还原结果表明,该系统能及时发现装配过程中的缺陷,对漏装、错装的识别准确率达到99. 5??以上,满足复杂电子装备装配过程零件装配状态检测的需求。

Complex electronic equipment has complex structural composition, multiple assembly processes, high precision requirements, high degree of coupling between electromechanical and hydraulic systems, and difficulties in online detection. If there are any assembly errors in the components, it will lead to rework. In order to achieve online detection of the quality of the assembly process, an image detection system based on machine learning is designed. Firstly, a movable image acquisition device and camera layout scheme is designed based on the size of the equipment, which can be applied to image acquisition requirements of different sizes of equipment. Secondly, the distortion problems existing in the acquired images are corrected, and the segmented images are stitched through image fusion. Finally, deep learning algorithms are used to achieve image recognition, and knowledge graph is used to restore detection results. The results show that the system can detect defects in the assembly process in a timely manner, with an accuracy rate of over 99. 5% for identifying missing and incorrect assemblies, meeting the needs of detecting the assembly status of complex electronic equipment parts.

TP271