期刊名称：Eastern-European Journal of Enterprise Technologies
印刷版ISSN：1729-3774
电子版ISSN：1729-4061
出版年度：2016
卷号：4
期号：3
页码：25-30
DOI：10.15587/1729-4061.2016.76103
语种：English
出版社：PC Technology Center
摘要：The study proves that it is necessity to develop and specify the procedures of using a synthesis of the methods of optical flow and background subtraction in the case of controlling objects of complex shapes on a changing background and in the presence of small moving objects that are not subject to tracking. Research in this area can yield significant results to automate the movement parameters control over objects such as railway transport.For the described conditions of practical use, it is most of all advisable to synthesize the classical Lucas-Kanade method of optical flow and the Horn-Schunck method for segmenting the frames and identifying control zones. The study describes the procedures of choosing the size of control zones and analysing a joint movement in these areas, which makes it possible to identify the movement of a cut even if the cut has been formed from different categories of wagons.The suggested algorithms were tested on the classification hump at Odesa – the Classifying Section station (Ukraine). The obtained quantitative characteristics of the accuracy of recognizing cuts show that the conditional probability of correct work of the suggested approach is 0.8332, compared with 0.44 in the case of the classical Horn-Schunck method under the same conditions.
关键词：video analysis;optical flow;background subtraction;monitoring the parameters of cuts’ movement;classification hump
