Video Processing Techniques for Real-Time Traffic Applications (original) (raw)

The procedure of traffic checking is done predominantly using radar detectors, loop detectors and manual observation. Traffic checking incorporates a lot of tough standards to be followed while guaranteeing that there are no congested driving conditions in the streets. The existing techniques and their shortcomings are briefly stated in the paper. The need of great importance is an insightful traffic observing framework that can decrease the manual exertion and increment effectiveness. One potential arrangement is the utilization of digital video processing strategies. The examination of clever traffic signal control framework has recently got significantly more importance because of the increase in automobile jams and accidents. This paper proposes video processing frameworks and techniques for detection of speed of the vehicle, presence of the vehicle and presence of helmet on the rider. The strategy for the most part lies in the internal working of deep neural networks and feature extractors. The proposed frameworks are relied upon to perform better than the current less-proficient systems by in any event a factor of five times. By bringing the proposed new developments and methods into the real world, the pace of congested roads and accidents can be diminished fundamentally Keywords: CCTV, IoT, HOG, SIFT, SVM I. INTRODUCTION Traffic congestion, incidents and traffic rules violation are a common problem in most large and medium sized cities. The traffic management systems that are currently in place make use of radar detectors, inductive loop sensors, infrared detectors, manual observation techniques where traffic police manually control the traffic on roads or video surveillance techniques which follow a semi-manual approach where CCTV cameras capture the traffic flow in real time which are analysed in a control room by the traffic police. These methods not only lack efficiency but also require a lot of manual effort. In case of detectors and sensors, installation and maintenance also is very hard. The need of the hour is an intelligent traffic monitoring system that can reduce the manual effort and increase efficiency. One possible solution is the use of video processing techniques. The cost of installing and maintaining such systems are very low because, all that is needed is a smart CCTV camera with an embedded platform with minimal computing capacity. Installation of such system allows for wide area monitoring and also causes very less disruption of traffic while installation or maintenance. If computing is done at the source, it will reduce the amount of data that needs to be transmitted to the central traffic control system and also prevents delay in decision making. However one important factor in this is the classification of processing that needs to be done online and in real-time and the processing that can be done offline. In this paper we first provide a literature review that talks about the related works done and common techniques used in the field. Then we provide a brief overview of the various applications where real time video processing techniques can be used to manage, monitor or control traffic and traffic related incidents and violations. This is followed by a discussion on the offline and online algorithms that can be used in traffic management systems. The rest of the paper provides a brief discussion about the challenges that exist that we might encounter followed by an overview of the results we obtained and conclusion.