Abstract: All times, amounts of stick-materials are counted by man. This method is inefficient and veracious. In order to improve efficiency, it is very important to excogitate one method to realize count amounts automatically. This paper puts forward an improved method of outline extraction by study to familiar outline extraction method. Eliminate the influence of filament and object area by outline extraction, partition, closed operation, fine, then get more exact outline curve and figure out outline’s roundness degree and stick-material’s radius which is an important parameter to recognize stick-materials automatically. The result of program shows that improved method of outline extraction can extract stick-material’s outline exactly.

Key words: Stick-material, Image processing, Image recognition, Outline, Characteristic, Chain-code

摘要： 一直以来，对棒线材进行计数都采用人工方法，效率低且不准确。为了提高效率，需要研究出一种棒线材自动识别方法实现棒线材的自动计数。该文通过对常见的轮廓提取法进行研究，提出了改进的轮廓提取法，通过对棒线材灰度图像的轮廓提取、分割处理、闭运算及细化处理等消除了细线及目标区域的影响，得到了更准确的轮廓曲线并计算出轮廓的圆形度，进而计算出棒线材的半径，棒线材的半径是实现棒线材自动识别的重要参数。程序的运行结果表明，改进的轮廓提取法能准确地把棒线材端头轮廓提取出来。

关键词: 棒线材, 图像处理, 图像识别, 轮廓, 特征, 链码