Real time recognition of pedestrian and vehicles from videos

This paper presents a new approach in recognition of moving objects captured by a surveillance camera. We have limited our area of study to the recognition of pedestrians and vehicles as it has ever increasing importance in the captured security surveillance as well as traffic monitoring systems. The primary phase of the method is the detection of moving objects using background subtraction and edge based subtraction. In the next phase, Speeded Up Robust Feature (SURF) of the moving object is extracted along with the height to width ratio. These features are used to correctly recognize the moving objects which then differentiate it to a pedestrians or vehicles. We have tested the performance of the system with sample videos as well as real time videos. The system shows a considerable recognition rate of 70% for pedestrians and 80% for vehicles. Statistical measures such as false discovery rate, recall and precision are used to measure the performance of the proposed system and 0.75% recall and 0.97% for precision has been obtained for pedestrians and vehicles.