KINERJA METODE MAHALANOBIS DISTANCE YANG DIBENTUK DARI DUA UKURAN PUSAT DAN DUA DISPERSI MULTIVARIAT (UNTUK UKURAN SIMILARITAS KLASIFIKASI IMAGE)

Abstract

Distance is a space that connects two points or two locations which can be calculated by length and time. Distance is used to measure the similarity of two objects (for example, an image object). An image can be considered to be similar to another image if the similarity size value is small. On the contrary, if the value of the similarity distance between the training object and the object being tested is large, the object can be said to be different or not. In this design, the image classification of Lakes, Forests and Settlements will be carried out by taking the Color feature using the Color Moment extraction method and the Texture feature using the GLCM (Gray Level Co-occurrence Matrix) extraction method and taking the method of calculating the distance between one data and another data that has High similiarity using the Mahalanobis Distance calculation method with two center sizes namely Mean and Median and three multivariate dispersions, namely the covariance matrix formed by the mean center value, the covariance matrix formed by the value of Median, and the covariance matrix formed by the value of Grand Median. From the research conducted, the performance results that can be considered for use are the Mahalanobis Distance with a median center size with a covariance matrix formed by the Median center value with an accuracy of 69.855% and a covariant matrix formed by the Grand Median center value with an accuracy of 69.565%. In this case the percentage is taken from testing images based on color characteristics using the Color Moment extraction method.