遮挡与缺失场景下屏幕缺陷视觉检测

Visual detection of screen defects in occlusion and missing scenes

为提高缺陷检测的智能化水平,解决复杂场景下缺陷检测难题,结合目标检测与目标匹配、图像差分技术,搭建一种遮挡与缺失场景下屏幕缺陷视觉检测系统.该系统包括手机屏幕检测、遮挡检测、缺失检测及屏幕内容变化检测4部分.使用YOLOv8n模型检测出图像中的手机屏幕位置,使用多台相机及检测框过滤方式检测出遮挡的手机屏幕,使用目标匹配算法判断手机屏幕是否存在缺失,再通过改进的图像差分算法判断手机屏幕显示内容是否发生变化.通过生产现场采集数据,并对数据进行标注、增强及过滤等处理,然后对目标检测模型进行训练与测试,所得手机屏幕平均检测准确率为96.8%.在生产现场部署并应用算法模型,通过对应目标匹配及差分计算,实现了多路视频同时处理下手机屏幕缺陷的实时准确检测.相关算法的应用既可有效减少人工成本又能提高检测效率.该检测技术可用于针对手机、电脑和电视等电子屏幕的质量检测.

To improve the intelligent level and solve the difficult problem of defect detection in complex scenarios,a visual detection system for screen defects in occluded and missing scenes is constructed based on object detection and matching,as well as image difference technology.The modules for mobile phone screen detection,occlusion detection,missing detection,and screen content change detection are established in this system.The YOLOv8n model is used to detect the position of mobile phone screens in images.Multiple cameras and detection boxes are used to filter out obstructed mobile phone screens.Target matching algorithms are used to determine whether there are any missing mobile phone screens,and an improved image difference algorithm is designed to determine whether the displayed content of mobile phone screen has changed.The data is collected at the practical production site,and then annotated,enhanced,and filtered.The target detection model is trained and tested,and the average detection accuracy of mobile screen is 96.8%.The deployment and application of developed model on the production site have been completed.Through corresponding target matching and differential calculation,real-time and accurate detection of mobile phone screen defects under simultaneous processing of multiple videos has been achieved.The application of proposed algorithm can effectively reduce labor costs and improve detection efficiency.The proposed method can be used for electronic screen quality detection in fields such as mobile phones,computers and televisions.