Abstract: To meet the requirements of the online CCD vision measurement system for large forgings by use of ray band produced by plane ray source, this paper proposes the pattern about the extraction of central curve in ray band based on image subtraction segmentation and noise removal by wide run-length. The pattern consists of three major steps named as image subtraction segmentation, noise removal, ray band repairmen and central curve extraction. The image subtraction segmentation is the core idea of the pattern. The noise removal is the key step, which comprises semi-thresholding, noise removal by wide run-length and discrete point removal by eight-connected pixels. The ray band repairing includes main part repairing and edge restoration by regional growth method. The experimental results show that the pattern has the satisfactory extracting effect.

Key words: image subtraction segmentation, noise removal by wide run-length, forgings measurement, central curve in ray band, regional growth, eight connected pixels

摘要： 针对基于片光源产生光条的CCD视觉大型锻件在线测量系统的特点和要求，提出求差分割与泛游程去噪提取光条中心模式。该模式主要分为求差、去噪、光条修复与中心提取三个主要步骤，其中：求差是模式核心思想；去噪是关键技术，包括半阈值去噪、泛游程去噪、八连通去离散点等处理；光条修复包括主体复原与边缘区域增长修复。实验结果表明，该模式具有较为满意的提取效果。

关键词: 求差分割, 泛游程去噪, 锻件测量, 光条中心, 区域增长, 八连通