摘要
指出Capson游程分析轮廓提取算法存在的两点不足:一是所提取轮廓在水平方向凹形部分存在偏差;二是未考虑合并情况会导致提取复杂图案轮廓时出现部分轮廓丢失现象.通过增加分叉点和交汇点以及建立两种未考虑合并情况对应的规则,分别对这两点不足进行改进.实验表明,改进后的算法能精确提取任意复杂图案的全部内外轮廓,在提取高游程平均压缩率图像轮廓时耗时较少.该算法能实现流水线式并行轮廓提取,减少线阵扫描相机应用中等待数据获取的时间,提高在线检测速度,且成功应用于高精度PCB线路板缺陷检测系统,实现高分辨PCB图像轮廓快速准确提取.
Two drawbacks of Capson＇s boundary extraction algorithm were investigated. The first is that distortion occurs at the concave portion of boundaries. The second is that the algorithm could fail to extract all boundaries of images with complex patterns due to the lack of consideration in two merging cases. The drawbacks were overcome respectively by introducing splitting points and merging points and by building the rules to the two merging cases. Experimental results show that the improved algorithm can extract all boundaries exactly for images with varieties of complex patterns and reduce the computation cost for the images with high run average compression ratio. Moreover, the pipeline parallel processing can be utilized to decrease the waiting time for acquiring image data in application of line-scanning camera, and was successfully applied in PCB defect inspection system to implement rapid and exact boundary extraction for high resolution PCB images.
出处
《深圳大学学报（理工版）》
EI
CAS
北大核心
2009年第4期405-410,共6页
Journal of Shenzhen University(Science and Engineering)
基金
国家高技术研究发展计划资助项目(2008AA8041205)
关键词
图像处理
模式识别
轮廓提取
游程编码
游程平均压缩率
image processing
pattern recognition
boundary extraction
run-length encoding
run average compression ratio
