Publication: Preliminary study in evaluation of unconstrained minimum average correlation energy (umace) filter using raw video data from running event.
Loading...
Date
2016
Authors
Zakaria, Muhammad Haziq
Journal Title
Journal ISSN
Volume Title
Publisher
Abstract
The purpose of the study is to evaluate the performance of UMACE filter for the data taken during a running event, Kelantan Open Run 2015. The images were taken during the running event were classified as raw data for uncontrolled environment. The performances of the filter were also compared to the images from controlled environment, where the lighting and the posture were kept almost the same. The images from controlled environment were taken from three volunteers. The evaluation process were done and observed using the peak-toside lobe ratio (PSR) value, where the method already is used in most of face recognition method. The threshold of the PSR value were set to certain value based on the range of PSR value to evaluate the specificity, sensitivity, precision and accuracy of the UMACE filter in the face recognition process. The evaluation process also used different number of training images from 2, 5 and 10 to find the optimal number of training images for the UMACE filter. MatLab software was used for the study. The results of PSR value using raw data from uncontrolled environment and from controlled environment are not consistent. Histogram equalization method used as an image enhancement method also showed inconsistent result. Thus, producing low value of sensitivity, specificity, precision and accuracy percentage in most of the evaluation process except for the process using two training images where the accuracy rate at the highest value between 79 to 80 %. Optimal number of training images used also showed inconsistent result. From the result, further studies are recommended in order to improvise the performance of the UMACE filter before it can be implemented in the running event. Another image processing method should be explored to increase the overall performance of the UMACE filter.