Methods of artificial intelligence in the tasks of information and measuring analysis and digital processing of quasistationary aerovideoimages
Vyacheslav Barhotkin, Evgeny I. Minakov, D. S. Kalistratov, Mikhail P. Kochetkov · 2016
Considered the problem of megacities traffic, offers an original technical solution for the automatic detection of the projection of asphalt pavement in the sequence of quasi-stationary aerial video photography, assesses the effectiveness of this solution in the task of statistical tracking of road traffic, proposed architecture of automatic measuring system on the basis of that solution, evaluated the prospects of such system. At the heart of the proposed technical solution is a combination of artificial intelligence methods and techniques of digital image processing. The main idea is in the automatic recognition of asphalt pavement projection in the video series. This saves time for image processing and enables statistical analysis of the traffic in real time. This is achieved by use of a digital low-pass filter (time filter) to the color pixel signals of images. After applying this filter for cars in the picture begins to emerge their filtration trail. Recognition algorithm detects such tracks and gradually forms the asphalt mask for video. So, after some time, cars will show where the road by themselves. After final recognition of asphalt pavement processing algorithm ignores the remainder of the frame, which leads to saving computation time. On perspective with this technical solution it is possible to construct an automatic measurement system for analyzing traffic flows in real time, and now developed solution allows in semi-automatic mode to calculate the number of cars and the average speed of traffic with less than 10 % errors.