Vehicle Density Analysis and Classification using YOLOv3 for Smart Cities

• Published in: 2020 4th International Conference on Electronics, Communication and Aerospace Technology (ICECA) pp. 980 - 986
• Main Authors: Rashmi, C R, Shantala, C P
• Format: Conference Proceeding
• Language: English
• Published: IEEE 05.11.2020
• Subjects: Automobiles; Cameras; darknet; Deep learning; opencv; Real-time systems; Streaming media; Trajectory; vehicle classification; Vehicle density; Vehicle detection
• DOI: 10.1109/ICECA49313.2020.9297561

Incorporation of the digital technologies in the surveillance of urban mobility such as monitoring the traffic density will help in improving the quantity of vehicles/arrangements to be provided for the public commutation, facility to be incorporated in reducing the traffic, infrastructure to be provided such as road widening, pedestrian path, over bridge, under pass etc., where traffic and transport is an issue. This can be implemented in the city, at which it is recognized to be developed as a smart city. The proposed research work analyzes the vehicle density using python-OpenCV and YOLOv3. Real time videos are recorded in four directions from Sony HD IP cameras in a designated area. Image frames from video sequence are used to detect moving vehicles. The background extraction method is applied for each frame which is used in subsequent analysis to detect and count all the vehicles. The blobs are detected for each vehicle which helps to track the vehicle in motion. The center of each vehicle with blob gives the count of vehicle based on the lanes considered. This work not only counts the vehicle in real time but also classifies the different vehicles using deep learning technique. YoloV3(You only look once) object detection system is used along with a pretrained model called darknet to classify the vehicle into different categories (bus, car, motorcycle etc). This deep learning method showed better classification and detection rate compared to blobs and morphological method used for counting the vehicles. Classification is shown for vehicle and also person classification is considered to analyze the percentage of people and vehicles. The analysis of percentage of vehicles is shown using pie chart.