摘要
提出一种新的结合面向对象分类方法进行同态滤波的薄云去除方法,该方法采用Ecognition软件,将无人机影像划分为有云区域和无云区域,并对无人机影像进行同态滤波,用滤波后影像的有云区域代替原影像的有云区域,无云区域保持不变。并对处理后的影像进行监督分类,再对整幅影像进行Wallis滤波匀光匀色处理,使得有云区域的色调、对比度和亮度与无云区域趋于相同。实验结果表明方法效果较好,能有效地消除或减弱薄云对无人机影像的影响。
In this paper,we propose a new thin cloud removal method combining object-oriented classification and homomorphic filtering.This method utilized Ecognition software to divide the UAV image into cloud area and cloud-free area,and performed homomorphic filtering on UAV images.Then,the pixels in the cloud area were replaced by the pixels in the cloud area after filtering,and the pixels in the cloud-free area remain unchanged.Finally,the processed images were performed by supervised classification,and the whole image was processed by Wallis filtering,which made the hue and contrast in cloud area and cloud-free area tend to be similar.The experiment result shows that the proposed method has good effect and can effectively remove or reduce the impact of thin cloud to the UAV images.
出处
《地理空间信息》
2021年第10期83-86,I0002,共5页
Geospatial Information
基金
中国地质调查局地质矿产调查专项基金项目(1212011301000,1212011220149)。
关键词
无人机影像
同态滤波
面向对象分类
去云
UAV image
homomorphic filtering
object-oriented classification
cloud removal
