Invariant Feature-Based Darknet Architecture for Moving Object Classification

Object detection and classification is important for video surveillance applications. Counting vehicles like cars, truck and vans is useful for intelligent transportation systems to identify dense and sparse roads, track loaded vehicles at the country borders. Even though many solutions such as appearance-based (Multi-block Local Binary Pattern) and model-based ((DATMO) algorithm) are proposed to classify the moving objects within the satellite images using machine learning and deep learning techniques, they either have over fitting problems or low performance. Hence these challenges have to be addressed during detecting and classifying the objects. Instead of training the classifiers with hand-crafted features, this paper uses neural network based object detection and classification to achieve promising accuracy better than the humans. Invariant feature concept is added to the existing Darknet Architecture of You Only Look Once (YOLO) and is combined with Faster Region-Based Convolutional Neural Networks (Faster R-CNN) to count the number of vehicles with different spatial locations. This combined model improves feature extraction step and vehicle classification process. The proposed system is tested on two benchmark datasets Cars Overhead with Context (COWC) and Vehicle Detection in Aerial Imagery (VEDAI) for counting the cars and trucks. Experimental results prove that the proposed system is better by 9% in detecting smaller objects than existing works.