A Smart Surveillance System for Pedestrian Tracking and Counting using Template Matching

Human Detection and tracking have become a focal area of research as it plays a major role in computer vision applications. Methods and equipment for the detection and tracking of humans are constantly being changed as advanced methodologies are being discovered. Even though this is a huge topic in the computer vision field, pedestrian detection in infrared images is relatively new. In this Research paper, we designed a new methodology for pedestrian detection, tracking, and counting in infrared images. The main objective of our research is to give machines a technique for the detection of pedestrians in real-time. Our proposed method of human detection detects and tracks pedestrians effectively and efficiently using the OCTVBS benchmark Pedestrian Infrared/visible Stereo Video Dataset. First, we used Template matching for Pedestrian detection. After detections tracking is done using Kalman filter. Lastly, headcount is acquired from head detections gained from applying Haar-like features. We evaluated our method on the OCTVBS dataset and gained detection accuracy of 91.04%, tracking accuracy of 92.47%, and count accuracy of 82.11%. The presented method should be useful for the monitoring of pedestrians in IR video surveillance.