摘要
针对目前手机拍照识别过程中难以识别大角度旋转和俯仰角度变化目标的问题,设计并实现了能够容忍图像旋转和扭转变化的改进型SIFT关键点匹配算法;该算法通过模拟目标任意多角度水平旋转和垂直扭转,提前计算多组SIFT特征;在手机拍照识别景观建筑时,匹配目标图像的SIFT特征与多组特征从而识别目标多个侧面;实验结果表明,该识别系统对任意角度拍摄的手机图像的识别准确率达80%以上,计算耗时仅10s,具有一定的实用性和推广价值。
This paper develops a novel image matching technology for object recognition via mobile phone cameras.The affine and non-rigid changes of object 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.Experiment shows that the system achieves 80% accuracy which cost less than 10s.The technology and methods adopted in the system are practical and worthy of using abroad.
出处
《计算机测量与控制》
CSCD
北大核心
2012年第2期451-453,共3页
Computer Measurement &Control
