Real Time Vehicle Detection and Colour Recognition using tuned Features of Faster-RCNN

Abstract

Being the most dominant part of the vehicle, colour anticipate vital role in vehicle identification. Thus, colour also plays significant part in Intelligent Transportation System (ITS) and can be very effective in various applications of ITS. In past, most of the work had done on colour recognition of vehicle are not able to achieve the high accuracy because they rely on hand-crafted feature i.e. Speeded Up Robust Features (SURF), Scale Invariant Feature Transform (SIFT) and Histogram of Oriented Gradient (HOG). In this work, we proposed a solution by utilizing one of the latest deep learning algorithm for the detection of vehicle and the classification of detected vehicles colour. Proposed methodology is based on the tuned features of Faster R-CNN and achieved the good results as compared to current state of the art techniques. In addition to that, this work is also contributes towards the dataset collection of related vehicles being used in Pakistan. Proposed method outperformed the previous works by achieving 95.31% accuracy on testing data. The robust results in terms of accuracy and the generation of dataset depicts the novelty of proposed technique in the literature.