Image processing applied in road traffic using "Data from Sky"

Abstract

Detection of vehicles in traffic to measure microscopic and macroscopic traffic parameters is the subject of long research in the field of road traffic engineering. This article presents the results of the application on traffic video images of the algorithm that is in the "Data from Sky" system. As part of the video image processing, we used as a part of image acquisition, a drone, positioned nearby Cluj-Napoca City. By uploading the video image on the online platform "Data from Sky" you get a fully automatic service for extracting traffic data from low altitude drone cameras records. The system is using as background Deep Learning. The file generated online can be exploited to obtain tangible results from the video such as: Traffic flow, Average and instantaneous speeds, classification on vehicle categories, accelerations and decelerations of each vehicle, distance travelled by each vehicle in the video and in case of very good image quality can be extracted license plates of the vehicles. Also, in case of monitoring in the urban area, the pedestrians who transit the area of the surveillance camera will be counted. In this case study, we focused on extracting as many traffic parameters as possible, to have an overview of the analysed area.