结合超像素和卷积神经网络的国产高分辨率遥感影像云检测方法

由于国产高分辨率卫星遥感影像波段少、光谱范围窄,导致传统云检测方法精度低。本文提出了基于卷积神经网络的高分辨率遥感影像云检测方法。首先采用主成分分析非监督预训练网络结构,获取待测遥感影像云特征;然后采用超像素分割方法进行影像分割;最后将检测结果影像块拼接,完成整幅影像云检测。试验效果评价表明,基于卷积神经网络的高分辨率遥感影像云检测方法不受光谱范围限制,云检测精度高,误判较少,适合国产高分辨遥感影像云检测。

Cloud detection for Chinese high resolution remote sensing imagery using combining superpixel with convolution neural network

Accurate cloud detection in Chinese high resolution remote sensing imagery is very low, due to the fact that there are few image bands and limited spectral range. In this study, high resolution remote sensing image cloud detection method based on convolution neural network is proposed. At first, in terms of network training, this paper obtains the feature of remote sensing images through using principal component transform and unsupervised pre-training network structure. Then it enters the image block into the network for cloud detection using the superpixel segmentation method for segmentation. Finally, the test image block is spliced, and the entire image cloud detection is completed. Experiments show that this method has high computational accuracy and the ranges of the spectral bands are not a limit.