摘要
针对工业零件含噪图像边缘检测,根据Canny算法原理,提出了一些改进策略,形成了一种矩形透镜最大梯度模边缘检测算法。采用中值滤波完成图像平滑,有效抑制了图像噪声;采用5×5邻域一阶偏导有限差分计算图像的梯度幅值,提高了边缘定位的精度;采用最大类间方差法(OTSU)求解了最优区域分割阈值,实现了边缘的自动检测。以磁环和极片工业零件图像边缘检测为例进行了实验,结果表明,该算法具有较好的去噪和边缘检测效果。
In order to detect the edges of noise image for industrial components, based on Canny algorithm principle, the paper presented the rectangle lens maximum gradient magnitude algorithm method with some improvement strategies. The image is smoothed with median filter to effectively restrain images noise. The gradient magnitudes are calculated with the first-order derivative within 5×5 neighborhoods to improve the precision of the detected edge position. The edges are detected automatically based on the optimal threshold obtained by OTSU. Experimental results show that the presented algorithm has better effect for denoising and edge detection. [ Ch,4 fig. 13 ref. ]
出处
《轻工机械》
CAS
2012年第4期77-80,共4页
Light Industry Machinery
基金
浙江省公益性技术应用研究计划项目(2011C31048)
国家自然科学青年基金(61103171)
杭州师范大学钱江学院科学研究项目(2011QJJL09
2012QJXS14)
