An Improved Mean Shift Using Adaptive Fuzzy Gaussian Kernel for Indonesia Vehicle License Plate Tracking

摘要

A new approach toward Indonesian vehicles license plate tracking based on video recordings of vehicles on the highway, is proposed. The tracking technique is used to improve the performance of a standard Mean Shift with a Gaussian kernel by selecting the appropriate kernel radius using an adaptive fuzzy mechanism. The purpose of kernel radius variation of Parzen window is to keep or maximize the mean of the similarity function outputs which implies a successful tracking process. The experimental results show that Improved Mean Shift using Adaptive Fuzzy Gaussian Kernel proved to have better effects as compared to the Standard Mean Shift.