Algorithms for Computer Vision Based Vehicle Speed Estimation Sensor

In this paper authors present a new computer vision algorithm developing and testing for vehicle speed estimation. After researching the literature, two methods of calculating speed have been proposed - sampling method and skipping method. Value of main parameter – number of measurement points - for each method has been tested and evaluated. The data obtained by these methods was compared with the reference speed data from the IR sensor group, the mean square error was calculated, and the analysis of the accuracy of these methods was carried out for various moving speeds. This paper is devoted to testing the previously developed system–computer vision-based sensor for detecting the position and speed of vehicles by using a single camera. The algorithm used by the sensor was verified and its error measured by using a stand with model vehicle driving around a known track. Infrared sensors were used to measure the real speed and compare with output of the developed system.