计算机工程与应用 ›› 2014, Vol. 50 ›› Issue (13): 168-172.

• 图形图像处理 • 上一篇    下一篇

采用轮廓向量特征的嵌入式图像匹配方法

倪  健1,白瑞林1,李  英1,吉  峰2,李  杜2   

  1. 1.江南大学 轻工过程先进控制教育部重点实验室,江苏 无锡 214122
    2.无锡信捷电气有限公司,江苏 无锡 214072
  • 出版日期:2014-07-01 发布日期:2015-05-12

Embedded image matching method based on vector features of contours

NI Jian1, BAI Ruilin1, LI Ying1, JI Feng2, LI Du2   

  1. 1.Key Laboratory of Advanced Process Control for Light Industry(Ministry of Education), Jiangnan University, Wuxi, Jiangsu 214122, China
    2.Xinje Electronic Co., Ltd., Wuxi, Jiangsu 214072, China
  • Online:2014-07-01 Published:2015-05-12

摘要: 为实现工业流水线快速、准确目标定位,提出一种采用轮廓向量特征的实时图像匹配方法。以[X、][Y]方向向量为描述的关键轮廓点集为匹配特征,根据模板具体信息,计算最佳金字塔分层数、模板旋转角度步长和缩放步长,图像金字塔最高层则利用二级筛选策略,依据待测图优先剔除大量目标非潜在位置区域,仅对剩余少量区域进行计算,非最高层则进行同步局部搜索图像区域构建及匹配。实际测试表明:对目标遮挡,光照变化,聚焦不准,对比度低等鲁棒性强,耗时为毫秒级,识别率达97%以上,可实现任意坐标、角度和缩放情况下的目标定位,可满足工业现场要求。

关键词: 机器视觉, 图像匹配, 关键轮廓特征, 图像金字塔, 仿射变换

Abstract: A real-time image matching method based on the vector features of contours is proposed to achieve quickly and accurately target location. The input is the [X/Y] direction vector features of the key contours. This paper computes the best constructed levels of image pyramid, gets the rotated angle step and scale step of template on every pyramid level, according to template image. On the highest level of image pyramid, it firstly removes positions where target exists impossibly. Then the paper constructs local region and locate target synchronously on the other levels. Experiments show that this method is robust against occlusion, nonlinear illumination, defocus, low contrast and so on. It needs some milliseconds and the recognition rate achieves 97% above. Target can be located at any coordinate, angle and scale, so it can satisfy industrial requirements.

Key words: machine vision, image matching, features of the key contours, image pyramid, affine transformation