An Improved Pedestrian Detection System that Utilizes the HOG-UDP Algorithm

There is a high level of interest in pedestrian detection systems based on worldwide acknowledgement of pedestrian safety, and the need for research in this area is increasing. The HOG based pedestrian detection method proposed by Dalal and Triggs has been recognized as being less sensitive to the clothing and poses of pedestrians and also changes in lighting and is therefore one of the main methods used for pedestrian detection. But because the HOG based method requires a significant amount of computations, it is difficult to implement this method in real-time. Therefore in this research study, to improve the speed and detection rate for pedestrian detection, a pedestrian detection method that was improved upon by reducing dimensions of the particular feature vector extracted using UDP dimension reduction was proposed, and the results of performance evaluation showed that compared to previous HOG, HOG-PCA and HOG-LPP etc. algorithms, the speed and detection rate of the proposed algorithm were confirmed to have been improved.