Pixels classification of cloud images and acquirement of whole-sky cloud images are two important aspects of ground-based cloud observation instruments. To improve the accuracy of pixel classification of cloud images and minimize distortion of whole-sky image,approaches of image segmetation in HSI colorspace based on mathematics expection and whole-sky image obtained by image mosaic based on non-project to plane transformation. Theory analyzing and experiments results demonstrate the effectiveness of approaches present here.%云点识别和全天空图像获取是地基光学全天空测云技术的两个关键问题。针对现有云点识别方法对薄云和淡云的识别偏少或漏检的情况,提出了饱和度空间基于数学期望的改进云图分割方法;针对现有全天空云图获取畸变较大的情况,提出了基于非水平面投影思想的改进图像拼接方法。通过对拍摄云图的云量进行估算与拼接处理实验分析可知,本文方法对于全天空云量估计更符合人工目测结果。
展开▼