Design of Wetland-ecological Corridor using Multi-scale Remote Sensing Image Segmentation Method

For the ecological environment and bio-diversity protection at Jiansanjiang area, China, thesis adopted multi-scale remote sensing image segmentation method to detect an ecological corridor between two national nature reserves Sanjiang(SJNR) and Honghe(HHNR). The width of this corridor was detected by calculating the division threshold value of NDVI standardized change intensity index. And the structure of corridor was designed using a moving window algorithm and a K-means cluster analysis algorithm. The results showed that: (1) the division threshold value was set at 20%, background noise was 9.43%, optimal width of wetland ecological corridor was 1298m. (2)Cluster cl, c2, c3 and c4 classes of 8 human disturbance intensity types were the weakest regions affected by human disturbance in the primitive ecological environment of Nong river, Wusuli river, Sanjiang nature reserve and Honghe nature reserve etc. To construct wetland ecological corridor, cl, c2, c3 and c4 respectively defined as 'core', 'test', 'edge', 'buffer' zone between two nature reserves. (3)Swamp at core zone accounted for 75%, the high accuracy was 93.7%, swamp at test zone accounted for 72.2% and the accuracy was 75.79%, margin and buffer had the effect of edge guardrail, buffer width was 945m. This study provided a reliable scientific reference for wetland ecological corridor construction and restoration.