Tree identification using a distributed K-mean clustering algorithm

Trees play an important role in maintaining environmental conditions suitable for life on the earth. To classify the tree type is very important for the forest maintenance. With the advent of high spatial resolution remote sensing sensors, our ability has greatly increased for tree type identification. Considering the amount of data in need of processing and the high computational costs required by image processing algorithms, conventional computing environments are simply impractical. Therefore, it is necessary to develop techniques and models for efficiently processing large volume of remote sensing images. In this study, a cluster computing environment was adopted to speed up the computation time. The test image was first partitioned into hundreds of manageable sub-images. Scheduled by the head node, the sub-images were then distributed to compute nodes for processing. A distributed K-mean clustering algorithm with undetermined number of class was applied to each compute node. A promising result was obtained. Compared to the field investigations, tree types of the test site were properly identified. In addition, great improvement in computation time was obtained. The distributed K-mean clustering algorithm implemented on our cluster computing environment performed much faster than stand-alone alternatives. By adding more compute nodes to our cluster computing environment, further improvement in computation time is expected.