Still image processing techniques for intelligent traffic monitoring

With the increase of urbanization, emerging cities all around the globe are trying to tackle their traffic woes by resorting to dynamic analysis that guides commuters and thus regulates traffic. There have been several attempts to carry out this analysis by video processing the live feed from surveillance cameras installed at junctions. In this paper we highlight the methods that could be used to detect the traffic density at a particular junction by processing a still aerial image. The techniques for image processing that are highlighted in this paper could thus be applied in situations where the surveillance camera is not fast enough to send the live video within a fraction of a second or in cases where the server of the website that displays the live feed from these surveillance cameras is slow. Also, installation of still-image cameras, rather than motion video cameras would be more cost-effective to emerging economies. We make use of the websites hosted by the Public Traffic Information System (usually found in most of the emerging cities) which have live traffic images taken from the surveillance cameras installed at various places. Most importantly, unlike the solutions given in several papers[3] that require sophisticated hardware installation, the solutions given in this paper could be applied with immediate effect as there are websites that give live feed as mentioned earlier. The traffic density is estimated by giving a rating on a scale of 10. We have proposed several techniques like temporal variance, finding the average distance between vehicles and pre-marking high density regions all of which could be implemented on an aerial image view of a traffic junction. Each of the image processing techniques we mentioned has a certain percentage of contribution to the final traffic rating(given on a scale of 10)