CLASSIFICATION BASED ON OBJECT ORIENTED ANALYSIS SEGMENTATION IMAGE SENSING FOR MUCH OF HIGH SPATIAL RESOLUTION

Remote sensing from satellites and aircraft rides, one of which produces the appearance of the earth's surface the data in detail for use in analysis or monitoring changes. Highly detailed image data (high spatial resolution) digital image processing method requires a more specialized object-oriented classification is based segmentation. Segmentation algorithms are pretty much going to raise the question, which is the most appropriate method is used for object-oriented classification for remote sensing image analysis of high spatial resolution. The purpose of this research is to create a system for classifying satellite remote sensing imagery with high spatial resolution of the object-oriented classification based on multiresolution image segmentation. Segmentation method used is KMeans Image Clustering, Fuzzy CMeans Image Clustering, KMeans Region Clusterer, Simple Region Growing, and Statistical Region Merging, Mean Shift. Six methods were compared with the classification accuracy of the results of field checks of data, to determine what the most appropriate method. The results showed segmentation method with Mean Shift algorithm has the highest accuracy compared to five other algorithms. When compared with classification without using segmentation (the original image), there is an increase up to 40.9% accuracy so it can be concluded segmentation method is an appropriate method for digital classification of high spatial resolution satellite imagery.