Background Subtraction Berbasis Self Organizing Map Untuk Deteksi Objek Bergerak

Abstract

First part in automatic video analisys is moving object detection. An accurate moving object detection is needed indeed to next step process of automatic video analisys like tracking object detected adn then analyze of detected object. Background Subtraction is a common approach in moving object detection. The common problems in background subtraction are illumination changes, object shadow, and dynamic background like waving tree. Self organizing Maps algorithm apllied in background Subtraction to handles these common problems. Median filtering and morphological operation added after background Subtraction procces in conjunction to increase and produce accurate moving object detection. Apllied SOM, median filtering, and morphological operation in background subtraction increasing object detection accuraccy with value of MSE in 1463,73 and PSNR in 17,035 compare with alpha based background Subtraction where 4268,50 for MSE and 12,018 for PSNR.