基于多目视觉的扭锁定位方法  

Orientation Method of Twist Lock Based on Multi-Vision

作　　者：张氢[1] 闵冠宇 覃昶 秦仙蓉[1] 孙远韬[1] 

机构地区：[1]同济大学机械与能源工程学院,上海201804

出　　处：《机电一体化》2023年第1期26-35,共10页Mechatronics

摘　　要：为配合港口集装箱扭锁装卸场景下的扭锁自动化识别,对扭锁视觉定位的问题进行了研究。通过实验研究了不同算法在模拟双目图像中的匹配效果,提出基于特征点匹配的扭锁视差测距方法,并且测试了该方法对扭锁测距的效果。基于传统的双目视觉方法,构建相应的多目视觉系统,通过两组双目相机的测量结果计算扭锁目标的位置,实现了扭锁目标在世界坐标系中的定位。最终解决港口装卸环境下对扭锁目标视觉识别中的定位问题,为后续硬件装置对扭锁的抓取操作提供了位置信息。For cooperation with the automatic identification of twist locks in the scene of port container stevedoring,the problem of visual positioning of twist lock was studied.Through experiments,the matching effects of different algorithms in simulated binocular images were studied.A twist lock parallax ranging method based on feature point matching was proposed,the effect of which on twist lock ranging was tested.Based on the traditional binocular vision method,the corresponding multi-vision system was constructed.The position of the twist lock target was calculated through the measurement results of two groups of binocular cameras,and the orientation of the twist lock target in the WCS was realized.Finally,the orientation problem in the visual recognition of twist lock target in the port stevedoring environment was solved,which provided position information for the subsequent hardware device to grasp twist locks.

关 键 词：视觉识别 港口集装箱 多目视觉定位 特征点匹配算法 

分 类 号：U65[交通运输工程—港口、海岸及近海工程] TP391[交通运输工程—船舶与海洋工程]