面向自动物流控制的多角度货品图像匹配研究

摘要

针对目前连锁超市等物流系统中商品类别控制的现有问题,采用先进图像匹配与控制技术开发了一种新型的连锁物流货品图像识别系统。该系统采用对旋转和尺度鲁棒的改进的图像SIFT关键点匹配算法,显著提高了任意角度拍摄的货品图像的匹配精度,提高了货品的识别准确率。应用结果表明,该识别系统运行稳定可靠,匹配精度高,具有一定的实用性和推广价值。%This paper develops a new type of image matching technology for product class recognition in supermarket logistic management. The scale and affine changes of product images are tackled by using the advanced SIFT keypoint matching algorithm. We increase the precision of matching where product images are taken from arbitrary angle and distance. The application shows that this system runs stably and accurately. The technology and methods adopted in the system are practical and worthy of using abroad.