Research On Cloud Detection Method Of High Resolution Remote Sensing Image

With the rapid development of remote sensing technology,high resolution remote sensing images have been widely used.In order to improve the efficiency of remote sensing image data transmission,this paper studies and designs a cloud detection method based on high resolution remote sensing image in view of the problem of cloud coverage occupying channel transmission bandwidth.In this paper,The purpose of cloud detection is transformed into the dichotomous problem of clouds and ground objects.The problem is mainly composed of the characteristic attribute space of samples,the classifier design based on support vector machines,the experimental verification of cloud detection algorithms and the spatial scale analysis.The construction of sample feature space is the basis of cloud detection.In this paper,cloud and feature attributes are analyzed from four aspects:physical causes,imaging links,texture attributes and fractal geometry.The gray mean value of cloud and ground objects,Feature and fractal geometry,and constructed the feature attribute space according to the weak correlation between features and the classification ability of cloud and feature.With the sample attribute eigenvalue and sample identification as input space,a classifier is designed based on support vector machine.In this paper,we combine the classification ability of cloud and ground objects and the cross-validation result of the classifier to iterate the sample multiple times,and ultimately construct a high-quality sample space,which fundamentally improves the performance of the cloud detection algorithm.In classifier training,the influence of the number of samples on the performance of different kernel functions is analyzed.After the classifier is obtained,the cloud detection algorithm in this paper is experimentally verified by using remote sensing images and ground experiments,and the performance of the cloud detection algorithm is analyzed.Experimental results show that the cloud detection algorithm proposed in this paper has a high early warning rate and a low false alarm rate,and has good detection capability for different underlying surface scenarios.Using the pre-warning index of cloud detection to evaluate the influence of different spatial scales of remote sensing images on the performance of cloud detection algorithm,and based on this,the range of spatial scale of remote sensing image is analyzed.