摘要： 针对基于阈值分割的多晶硅晶畴检测方法中出现的边缘断点以及伪边缘问题,提出一种基于空间梯度信息的自适应边缘提取优化算法。对双峰法和Otsu法相结合的图像分割算法进行改进,使分割阈值能够根据图像特征自适应地选取。结合原图的空间梯度信息得到梯度边缘,将梯度边缘与初始边缘进行综合处理以连接边缘断点并去除伪边缘。实验结果表明,该算法简单有效,对不同特征的多晶硅图像均能较好地实现自适应提取完整连续的边缘。

关键词: 多晶硅, 晶畴识别, 图像分割, 边缘优化, 自适应检测

Abstract: In order to solve the edge breakpoint and pseudo edge problem occurred in the detection of polycrystalline silicon crystal domain based on threshold segmentation,an adaptive edge extraction optimization algorithm based on spatial gradient information is proposed.The image segmentation algorithm combining bimodal method and Otsu’s method is modified,so that the segmentation thresholds are selected adaptively according to the image characteristics.The gradient edge is extracted by using spatial gradient information of the original image.The edge breakpoint can be connected and the pseudo edge can be removed by combining the gradient edge and initial edge.Experimental results show that the proposed algorithm is simple and effective.Completed and continuous edges can be extracted adaptively for polycrystalline silicon images with various characteristics.

Key words: polycrystalline silicon, recognition of crystal domain, image segmentation, edge optimization, adaptive detection

中图分类号: 

• TP301.6