Gaussian Mixture Models Optimization for Counting the Numbers of Vehicle by Adjusting the Region of Interest Under Heavy Traffic Condition

Mixture models research has been widely implemented for numerous purpose in motion tracking applications. This method usually applied for tracking and counting vehicles in Intelligent Transport System (ITS). In this context, Mixture Model chosen is Gaussian Mixture Model (GMM) method due to its powerful features. Unlike many motion-tracking based methods, GMM achieves satisfactory performance from its ability to handle background subtractions. However, its implementation in detecting vehicle still have unsatisfactory result in accuration and identifying object, mainly under heavy traffic condition. The problem turn to poor accuration of object detection. Therefore, in this paper, we propose optimization of GMM performance by adjusting the Region of Interest (ROI). The propose technique to completing the report by compare the result before and after experiment in separate condition. The result show that this approach leads to improvement in tracking and counting average of accuration of motorcycle by 6.97% and car by 39.04% in several condition. Our approach to modified the method has been experimentally validated showing better segmentation performance, and this is an unbiased approach for assessing the practical usefulness of object detection methods for vehicle under heavy traffic condition on the highway