Abstract: To solve the factors of adverse environments and poor shooting conditions which cause terrible impact to the rapid and accurate identification of DPM bar code from complex images, the significant amount of right-angle corners of the images and the inner characteristic of the existence of bright and dark corners being taken into account, it proposes a novel self-adaption corner detection algorithm based on the mask algorithm theory. The algorithm uses the smaller mask, which will just sort out the corners that may be the right-angle vertexes and use circle mask as a filter, the characteristics of the DPM images are considered when choosing the shape and the range of the mask, the threshold of local binarization as well as the criteria of the filters. Experimental results indicate that in the sample of 160 DPM images, the even accuracy is 88%. The algorithm has the ability of detecting the corner of the bar code from all kinds of images and eliminating most of the non bar code corners.

Key words: corner detection, Direct Part Mark（DPM） bar code, self-adaption threshold, complex condition

摘要： 为了解决工况环境恶劣和拍摄条件差等因素导致的从复杂条件图像中快速准确地检测出DPM条码区域的困难，抓住DPM条码图像多直角角点及亮暗两类直角点共存等内部特征，提出一种新的基于掩膜思想的自适应角点检测算法。该算法使用小掩膜筛选出仅可能是直角顶点的那些点交予大的环状掩膜进行再过滤，掩膜形状和尺度、局部二值化阈值及相关的系列过滤条件的选定充分融入了DPM区域知识。实验结果表明，在样本为160幅DPM图像的实验中，DPM角点平均检出精度为88%。所提出的算法在面对复杂条件下的DPM图像时，有能力在检出DPM区域角点的同时抑制掉绝大多数非DPM区域角点。

关键词: 角点检测, 直接零件标识（DPM）条码, 自适应阈值, 复杂条件