基于阈值递归选取和引导滤波的航拍图像云检测

Cloud Detection of Aerial Images Based on Threshold Recursive Selection and Guided Image Filter

为了检测出航拍图像中的云区,提出一种基于阈值递归选取和引导滤波的云检测算法。首先基于HSI色彩空间构建云区显著图,拉伸云区和非云区的差异;再根据最大类间、类内方差比法进行阈值递归选择,综合云区统计特征得到初步云检测结果;然后利用引导滤波优化薄云和半透明云检测,最后通过形态学闭运算得到最终的云检测结果。对比试验表明,该算法的结果具有更低的错误率,更符合人的视觉检测效果。

In order to detect cloud region of aerial image,a cloud detection algorithm based on threshold recursive selection and guided filter is proposed.Firstly,a cloud area salient map is built based on HSI color space.The effect is to stretch the difference between cloud area and non-cloud area.Then the threshold selection is based on the recursive maximum inter class and intra class variance ratio method,and the preliminary cloud detection results are obtained by integrating the statistical characteristics of cloud region.After that,guided image filter is used to optimize thin cloud and translucent cloud detection.Finally,the final cloud detection results are obtained through morphological closed operation.Comparative experiments show that our algorithm has lower error rate and is more suitable with human visual inspection.