Vehicle Detection and Classification via YOLOv4 and CNN over Aerial Images

Advanced traffic monitoring systems face major vehicle detection and classification challenges. Conventional methods need significant computational resources and struggle to adapt to various data collection methods. In this study, the presented method is an inventive way to classify and identify cars in aerial images. The model suggested has five phases. Initially, in pre-processing, the techniques are employed by eliminating noise unravelling through Contrast Limited Adaptive Histogram Equalization (CLAHE). Secondly, fuzzy C-means segmentation is used to differentiate the body setting and then detection of vehicles is achieved by YOLOv4. Finally, feature extraction utilized the orientated FAST, Rotated BRIEF (ORB), and Scale Invariant Feature Transform (SIFT) to differentiate between vehicles. Consequently, the classification of things categorically to several classes utilized a convolutional neural network (CNN). Having experimented with the model on several datasets that include the Vehicle Aerial Imagery from a Drone (VAID) and Vehicle Detection in Aerial Imagery among others, we have established that our model achieves a better performance. On VEDAI, the model achieves 96.1% operational efficiency while on VAID the model achieves 96.8% operational efficiency. Consequently, our methodology outperforms some of the existing methodologies used in the industry.