基于双目立体视觉的物流包装箱尺寸测量研究  

Dimensional Measurement of Logistics Packaging Box Based on Binocular Stereo Vision

作　　者：张志刚[1] 霍晓丽[1] 周冰[1] ZHANG Zhi-gang;HUO Xiao-li;ZHOU Bing(School of Information Technology,Jiaozuo University,Jiaozuo 454003,China)

机构地区：[1]焦作大学信息工程学院,焦作454003

出　　处：《包装工程》2020年第19期230-236,共7页Packaging Engineering

基　　金：河南省科技厅软科学项目(182400410073);河南省高等学校青年骨干教师培养计划(2018GGJS267)。

摘　　要：目的针对物流行业运输中包装箱的非接触自动测量,提出一种基于改进SURF配准算法的双目立体尺寸测量方法。方法首先采用二进制FREAK描述子代替传统SURF的描述子,解决传统SURF描述子计算耗时、描述向量生成依赖于特征主方向,且主方向计算误差会在后续步骤中出现传导放大的缺点;其次,采用PROSAC删除误匹配点,并利用FREAK级联匹配的方式进一步提高算法的匹配速度和匹配准确率。最后,利用视差优化和边缘提取算法获得精确三维空间体,实现非接触尺寸测量。结果实验表明改进算法可快速提取图片特征点并准确匹配,对不同规格包装箱检测结果显示,基于改进算法的测量方法测量误差小,检测速度快。结论改进图像匹配算法可有效提高图像匹配准确率,减少测量时间,对于提高物流行业运输效率、减少人工成本具有重要意义。The work aims to propose a binocular stereo size measurement method based on the improved surf registration algorithm for the non-contact automatic measurement of packaging box in the transportation of logistics industry.Firstly,binary FREAK descriptor was used to replace the traditional SURF descriptor,which solved the shortcomings of calculation time-consuming,the generation of description vector depending on the main direction of feature,and the main direction calculation error enlarging the conduction in the following steps in the traditional SURF descriptor.Secondly,PROSAC was used to delete the mismatching points,and the FREAK cascade matching method was employed to further improve the matching speed and accuracy of the algorithm.Finally,the accurate 3D volume was obtained by disparity optimization and edge extraction algorithm to realize non-contact dimension measurement.The experimental results showed that the improved algorithm could quickly extract image feature points and match them accurately.The detection results of different packaging boxes showed that the measurement method based on the improved algorithm had small measurement error and fast detection speed.The improved image matching algorithm can effectively improve the image matching accuracy and reduce the measurement time,which is of great significance to improve the transportation efficiency and reduce the labor cost in the logistics industry.

关 键 词：双目立体视觉 图像匹配 尺寸测量 物流行业 

分 类 号：TB485.3[一般工业技术—包装工程] TP391[自动化与计算机技术—计算机应用技术]