Vehicle Detection and Car Type Identification System using Deep Learning and Unmanned Aerial Vehicle

Abstract

Background/Objectives: UAV (unmanned aerial vehicle) based traffic measurement system has various advantages than the traditional traffic monitoring systems using fixed loop sensors. This paper proposes the designing and implementing method of vehicle detection and type identification using Deep Learning and UAV in UHD (ultra-high definition) 4K video images. Methods/Statistical analysis: The present study proposes the implementation method of detection and classification system that can be accomplished vehicle classification according to AUSTROADS�s plan. The proposed system has designed two primary processes: detection and classification. This paper introduces the method that vehicle data training, detection, and classification method by applying a Darknet-53 for vehicles found in UHD images. Also, we propose the variable classification method due to parked and stopped cars for traffic flow monitoring. So, we considered the three conditions of driving, stopping, and parking. Findings: The results of the experiment show that the proposed approach resulted in errors that were twice as low as conventional methods that are using a fixed search area. Improvements/Applications: The proposed study can be applied to traffic flow monitoring system, ITS (intelligent transport system), vehicle detection and classification system.