Surveillance systems are to be effective in current scenario, for which development is the challenging task for researchers. The major challenge is pedestrian detection in various condition of light. This can well assist to the advanced driver system. In this paper, an approach is considered for pedestrian detection and tracking. It can avoid the collision among vehicles and pedestrians as well. The different condition of light is considered as in day, night, and fog. Therefore a dataset is developed also. The results is compared with the standard multispectral dataset KAIST. Experimental results show the performance as high as possible.