A snow cover range extraction algorithm based on multi-temporal GF-4 satellite data
WU Wei1， LIU Yu2
1.National Disaster Reduction Center of China， Ministry of Civil Affairs， Beijing 100124， China；2.School of Remote Sensing and Information Engineering， Wuhan University， Wuhan 430079， China
Abstract：Snow cover monitoring is the foundation of earth science research， which has important significance for studies， such as global climate change， disaster prevention and disaster relief work. Gf-4 satellite is the only civil high orbit satellite in national high resolution earth observation system major project， and it is also the world's first geostationary orbit high resolution optical imaging satellite， which has broad application prospects in the snow remote sensing monitoring with its flexible and high frequency observation ability. In this paper， a method of snow cover range extraction is proposed based on multi temporal GF-4 satellite images. Using the advantage of GF-4 satellite such as multi spectrum， high efficiency， large swath and mid-resolution imaging， and combining reflection characteristics of snow and moving clouds in multi-temporal images， this method uses three main steps to extract the snow cover range. First， panchromatic images are processed by binarization segmentation to eliminate influence of low reflection targets. Then， the multi-temporal cloud and snow coverage information are synthesized to remove influence of moving clouds. On the basis of the above， by using the difference of the reflection value change of the cloud and snow in multi-temporal near infrared images， the multi-temporal minimum value synthesis and threshold segmentation of GF-4 satellite infrared images are carried out to refine snow cover range， which can further remove the influence of changing cloud. Taking the central and western regions of Xinjiang as experimental regions， the conclusion are drawn as follows through comparing the results of snow cover based on HJ-1B satellite. The spatial distribution trend of snow cover using multi-temporal GF-4 satellite data is consistent with that using HJ-1B satellite data. The snow cover monitoring region using GF-4 satellite is more extensive， and the overall accuracy of information extraction using GF-4 satellite images is 92.19%， which is higher than that of 89.84% using HJ-1B satellite images. However， due to the impact of the spatial resolution and the "constant cloud"， the snow recognition accuracy using GF-4 satellite images is 85.16%， which is lower than that of 94.53% using HJ-1B satellite images.
2018, Vol. 52 
