Detecting and Counting Pedestrians in a Crowded Environment Using ROI Mask in a Video

Pedestrian counting plays an important role in public safety and intelligent transportation systems. The techniques of tracking a single pedestrian becomes impractical and complicated, when the scenes are densely crowded. In this paper, first step is extracting foreground objects from the background by applying ROI mask and then the pedestrians are detected by segmentation method. Second, the large sets of features extracted are reduced by using SLFs in order to attain high accuracy for counting number of pedestrians. The bounding box representation is used for counting number of pedestrians in a crowded environment. The number of people can be estimated by using connected component labeling method. Finally, by using above features reliable people counting in crowd environment can be achieved. The proposed method is robust for illumination changes and working well in public places for safety of the people, such as railway station, hospitals, shopping malls, etc. For further implementation, consider the proposed approach on a dataset consisting of a large number of people and study how to refine the computational efficiency for large scale datasets.