CCTV와 IR카메라를 융합한 적응적 보행자 추적 알고리듬

Abstract

In this paper, an adaptive pedestrian detection algorithm using the fusion of CCTV and IR camera is proposed. One of the drawbacks of the pedestrian tracking algorithm using CCTV is the inability to track the object in the absence of light. While in the IR camera , the light is spread in the short distance. The proposed method is aimed to overcome these drawbacks. In CCTV, the pedestrian tracking method uses Haar-like Feature. The advantage of Haar like feature is the rapid computation of the feature values. A learning algorithm, based on AdaBoost is implemented, which selects a small number of critical visual features from a larger set and yields extremely efficient classifiers. The area of a pedestrian rectangle is estimated based on the size of the object. Similarly, In IR camera, the IR captured video is converted into binary video. Then the contour based tracking of body parts is employed. In this step, the tracking is performed separately for each part of the body and then the object rectangles are grouped into a single rectangle. The average value of the CCTV pedestrian rectangle and the IR pedestrian rectangle is calculated. Finally, the new rectangle is used by both CCTV and IR tracking algorithms.