CAJ | 학술논문

晶畴的区域分割提取是利用数字图像实现多晶硅结晶度检测中的一项关键技术．其中阈值的自动选择是实现自动化检测结晶度的核心。针对多晶硅图像的特点．充分利用图像的直方图信息．通过结合双峰法与波形对称性分析首先得到阈值数以及阈值的初步取值，然后利用最大类间方差法精确确定闽值．再进行硅片图像分割。实验证明．这种方法结合了双峰法和最大类间方差法的优点．在对多晶硅图像进行晶畴分割提取时能够获得优良的检测效果并明显提高算法的效率．为实现工业在线自动检测奠定坚实的基础。
정주적구역분할제취시이용수자도상실현다정규결정도검측중적일항관건기술．기중역치적자동선택시실현자동화검측결정도적핵심。침대다정규도상적특점．충분이용도상적직방도신식．통과결합쌍봉법여파형대칭성분석수선득도역치수이급역치적초보취치，연후이용최대류간방차법정학학정민치．재진행규편도상분할。실험증명．저충방법결합료쌍봉법화최대류간방차법적우점．재대다정규도상진행정주분할제취시능구획득우량적검측효과병명현제고산법적효솔．위실현공업재선자동검측전정견실적기출。
Digital image based domain segmentation is one of the key techniques for crystallinity detection of polysilicon. Automatic determination of thresholds is the core task to realize automatic machine detection. Based on histogram information and the characteristics of image, uses bimodal method and wave symmetry analysis to get the number of thresholds and their probable values, then utilizes Otsu＂ method to obtain the precise values of the thresholds. The experiment shows that this method combines the advantages of bimodal method and Otsu＂ method. Segmenting polycrystalline image with this algorithm can get well detection results and time consumption of this algorithm and has reduced to 1/3 - 1/4 compare with traditional. This method provides a solid foundation for the industrial automatic-line detection.