Algorithm 2: Distributed construction algorithm of image pyramid model.

Input: $RS$ is the source of RS images; $cols$ and $rows$ are number of rows and columns respectively; $CellSize$ is a function to get pixel value; $tileSize$ is the slice size of image tile.  Output: $AttributeData$ is the attribute data; $NetData$ is net data; 1:Read the partitioned image $RS$ into HDFS and achieve the coordinate range of RS images $Extent$; 2:Evaluate pixel value of RS images $cellSize=CellSize(Extent.width/cols,Extent.height/rows)$; 3:Divide layers using $MDLA(Extent,tileSize,cellSize)$ in the Algorithm 1; 4:Evaluate image resolution $Re{s}_i$ when dividing each layer i, and produce division data $divData$; 5:If the terminal condition of MDLA is satisfied, return final number of division layers $level$ and then implement Step 7; 6:If the terminal condition of MDLA is not satisfied, then $i=i+1$, and return Step 4; 7:Ingest the data of current layer to HDFS with two parameters: i, $divData$; 8:Based on two parameters in Step 7, call $SinkLevel$ function to establish the pyramid HDFS model, and produce the attribute data $AttributeData$ and net data $NetData$; 9:If $level\ge i$, repeat Step 8. If $level<i$, the repeat stops, $AttributeData$ and $NetData$ are outputted into the $HDFS$ finally;