Examples major classes. These samples are evaluated based

Examples of satellite image segmentation using SOMs-HyGATo demonstrate the practicality, strength, accuracy , and effectiveness of the SOMs-HyGA an experiment on a Spot 4 satellite image is implemented (Figure 7a), andat the end of the segmentation processes (Figure 7b) a number of samples are selectedfrom the segmented images which represent major classes. These samplesĀ are evaluated based on a field work accompanied with advanced geospatial technologiessuch as Global Positioning System (GPS). Later confusion matrices 47 isused to compute the accuracy . The matrix consists of information about actual andpredicted results which are created by the field samples collection and the segmentationmethod. Performance of such systems is commonly evaluated using the datain the matrix. This task is completed with four classes 1 = Crop 1 (light green), 2 =Infrastructure/ Urban (light brown), 3 = Shrub (light pink), 4 = Crop 2 green) withhundreds survey points (see Tables 3). The accuracy of SOMs-HyGA is 92%.Matrix of Spot 4 satellite image segmentation by SOMs-HyGASpot 4 satellite image (a) Original (b) Segmented by SOMs-HyGAThe second experiment is implemented on different type of satellite image theIKONOS image (Figure 8a) that has high spatial resolution of 1 meter. The SOMs-HyGA is used to segment the IKONOS image with high quality (Figure 8b) whichimproves clearly this important step of image processing. Moreover, the confusionmatrix is used to prove the high accuracy of the results (Table 4).Matrix of IKONOS image segmentation by SOMs-HyGAIKONOS (a) Original image (b) Segmented by SOMs-HyGAFour classes are used and evaluated with hundreds of distributed samples collectedin the field. These classes are the following: 1- Water bodies (blue to darkblue), 2- Vegetation (green), 3-Bare land (grey), and 4- Rock (white). The confusionmatrix shows that the accuracy of SOM-HGA can reach 95%.