Driver Assistance for Safe Navigation Under Unstructured Traffic Environment

This research work aims at developing a system using a vision sensor, which aids in detecting and warning driver about road anomalies on unstructured traffic environment like jaywalkers, two wheelers and other vehicles whose sudden entry into the drive region causing mishaps due to lack of attention by the driver while driving in unstructured traffic. Drive region on the road is predicted based on vanishing Point (VP) which is calculated based on camera calibration. AdaBoost machine learning algorithm with Local Binary Pattern (LBP) features is used to detect road vehicles. To remove false positives and improve detection and classification accuracy, Histogram of Oriented Gradients (HOG) features with Radial Basis Function (RBF) Support Vector Machine (SVM) is been developed. Visual alert of the type of the object that is entering and is in the drive region within the distance threshold of 15 meters from the vehicles is been developed and tested on road at accuracy of 0.73 F1 score.