Short term prediction of statistics for bigdata in video surveillance

• Main Authors: Kannikar Intawong, Kitti Puritat, Piyapat Jarusawat
• Format: Conference Proceeding
• Published: 2019
• Subjects: Computer Science; Engineering; Mathematics
• Online Access: https://www.scopus.com/inward/record.uri?partnerID=HzOxMe3b&scp=85066505513&origin=inward; http://cmuir.cmu.ac.th/jspui/handle/6653943832/65506
• Institution: Chiang Mai University

© 2018 IEEE. This paper is focused of traffic videos. Now a day, large amount cameras are installed in cities for automatic processing. The objectives of this work is to help the traffic expert to take decisions in real time such as accidents, congestion, etc.), or to schedule works to improve the traffic calming, for example to prevent an excessive speed or to build additional lanes. We compute statistics throughout the day and the week. The video analysis face the large difficulties such as illumination changes or occlusions. Our approach considers objects detection and objects tracking. In these problems, we try to make the robust systems for individual tracking stage. Additional, we predict the statistics by deep learning LSTM and compare with the mechanic flow method, which obtain a global information on the flow of objects in the scene.