用于机械系统瞬时目标的双视角高速视觉检测系统

发布时间：2018-06-01 09:43

本文选题：计算机视觉 + 高速视觉　；参考：《光学精密工程》2017年10期

【摘要】：基于高速相机研发了双视角高速视觉检测系统。该系统可从不同角度对周期机械运动进行长时间、无接触的同步实时检测,并可自动判断和存储异常动作的瞬时高帧率视频图像。基于机械系统运动特性提取理论,提出了多视角高帧率视频图像的离线处理算法,利用两个高速相机准确建立了样本图像数据库及它们的对应关系。提出了多视角同步实时检测算法,完成了对目标的实时视觉检测,并实现了瞬时保存关键图像的功能。为验证算法的有效性和系统的稳定性,搭建了双视角高速视觉检测系统,对真实的机械系统进行了实验。结果显示,提出的系统实现了500fps的双目高速相机同步检测,以及0.1s内运动目标的自动判断,同时完成了对关键图像的瞬时保存。文中的研究为机械系统的非接触检测提供了一种高速可视化的检测方法。
[Abstract]:A two-angle high-speed vision detection system based on high-speed camera is developed. The system can detect the periodic mechanical motion from different angles for a long time and without contact, and can automatically judge and store the instantaneous high frame rate video image of abnormal motion. Based on the theory of mechanical system motion characteristic extraction, an off-line processing algorithm for multi-view high frame rate video images is proposed. The sample image database and their corresponding relations are established accurately by using two high-speed cameras. A multi-view synchronous real-time detection algorithm is proposed. The real-time visual detection of the target is completed and the function of instantaneous saving key images is realized. In order to verify the validity of the algorithm and the stability of the system, a high speed vision detection system with two angles of view is built, and the real mechanical system is tested. The results show that the proposed system realizes the synchronous detection of the 500fps binocular high-speed camera and the automatic judgment of the moving target within 0.1 s. Meanwhile, the instantaneous preservation of the key images is accomplished. The research in this paper provides a high-speed visual detection method for non-contact detection of mechanical system.
【作者单位】：北京交通大学机械与电子控制工程学院;
【基金】：科技部国家重点研发计划资助项目(No.2016YFB1200100) 中央高校基本科研业务费专项资金资助项目(No.M16JB00240)
【分类号】：TP391.41

【相似文献】