图像驱动的立方体顶视图感知定位  

Image-Driven Top View Perception Positioning of Cubes

作　　者：李昂龙 孙轶南 陆思良[1] 王骁贤[3] 宋俊材 LI Anglong;SUN Yinan;LU Siliang;WANG Xiaoxian;SONG Juncai(School of Electrical Engineering and Automation,Anhui University,Hefei 230601,China;School of Internet,Anhui University,Hefei 230601,China;Department of Precision Machinery and Precision Instrumentation,University of Science and Technology of China,Hefei 230027,China)

机构地区：[1]安徽大学电气工程与自动化学院,合肥230601 [2]安徽大学互联网学院,合肥230601 [3]中国科学技术大学精密机械与精密仪器系,合肥230027

出　　处：《组合机床与自动化加工技术》2025年第3期55-60,66,共7页Modular Machine Tool & Automatic Manufacturing Technique

基　　金：国家自然科学基金项目(52375522,52075002,52207036);安徽省科协青年科技人才托举计划项目(RCTJ202421)。

摘　　要：提出了一种基于动态规划算法的最大内接矩形定位技术,用于实现基于机器视觉的拼图机器人对立方体抓取定位精度提升。所提方法通过单目相机获取目标立方体的图像,经过对图像进行预处理,包括阈值化和检测分割,并在二值图像中计算内接矩形,能够获得立方体顶面的矩形范围,计算出抓取的实际位置,进一步提高抓取立方体物品的精度。经过实验验证,算法具有高精度和高速度的性能,对于立方体物品的视觉定位,相比于其它的内接矩形算法,将定位的位置误差降低到1 pixel以下,平均误差仅为物体尺寸的0.38%,可以有效提升拼图机器人定位精度。This paper proposes a maximum inscribed rectangle positioning technique based on dynamic programming algorithm,aiming to improve the positioning accuracy of cube grasping for puzzle robots based on machine vision.The proposed method obtains the image of the target cube through a monocular camera,and after preprocessing the image,including thresholding and detection segmentation,it calculates the inscribed rectangle in the binary image,which can obtain the rectangular range of the top surface of the cube,and calculate the actual position of the grasped object,thereby further improving the accuracy of grasping cube items.After experimental verification,the algorithm has high precision and high-speed performance.For the visual positioning of cube items,compared with other inscribed rectangle algorithms,the positioning error is reduced to less than 1 pixel,and the average error is only 0.38%of the object size,which can effectively improve the positioning accuracy of puzzle robots.

关 键 词：机器视觉 内接矩形 动态规划 图像处理 拼图机器人 

分 类 号：TH165[机械工程—机械制造及自动化] TG659[金属学及工艺—金属切削加工及机床]