Abstract

Application of image processing to real-time systems is growing at a rapid speed and has great potential for exploitation. Human errors are inevitable and a system designed with predictable response time and a deadline, aid in overcoming these shortcomings. This paper implements a pedestrian and vehicle detection system that combines Histogram of Gradients (HOG) and Support Vector Machine (SVM) with a Cascade Classifier and Haar. The implementation described processes every 10th frame of the video and is tested on standard datasets. Error percentage was calculated using Weak law of large numbers for the algorithms and the improvised result after preprocessing the image. Our approach extends the functionality of the inbuilt people detector and trains a classifier to detect 4 wheelers, it minimizes processing delay and errors by tweaking the region of interest as per the requirement of a real-time application.