The Use of Gaussian Mixture Model for Counting Human Object on Video

Abstract

Video monitoring has been widely used in various places, for example, tourist attractions, stations, terminals, offices, supermarkets, minimarkets, and other places. The use of video monitoring has the aim to improve security aspects in a place that is considered quite helpful. The rapid development of technology, video monitoring has been implemented for purposes other than for security, such as people counting systems. People counting to find out the number of visitors in a place or building is a difficult job to do, requires a lot of time and often the data obtained is not appropriate. Gaussian Mixture Model (GMM) is a background reduction method, which is used to identify the background and foreground. Background reduction is an approach that is widely used to detect moving objects on video from static cameras. While blob detection is detecting a group of pixels connected in an image that has a different colour (white or black). Blob is used for object classification, whether the object is a person or not. From the results of system testing shows that counting people can be done well, with an average percentage of 95.83% recall, 94.5% precision and 90.88% accuracy of the entire video test data. The use of GMM in counting people in this study can also count people who carry objects properly as long as they are not drawn as in the morning test video. In addition, this system can store the calculated data.