ANALYSIS OF WORK OF TECHNICAL VISION ALGORITHMS IN THE TASKS OF TRAJECTORY MEASUREMENTS
Abstract
The results of the analysis of correlation-extreme algorithms and algorithms for selecting ob- jects by color characteristics for solving problems of trajectory measurements by means of technical vision are presented. The parameters of noise immunity and speed are investigated. Methods for optimizing the structure of these algorithms for working in a complicated interference situation are proposed.
Recently, there has been a significant increase in the number and scope of unmanned aerial vehicles and a variety of transport robots. The transport load on urban infrastructure is significantly increased. This requires the creation of effective intelligent transport systems. A perceptible role in the structure of these systems is played by the means of optical motion monitoring and oriented to evaluate the trajectory parame- ters of the observable moving objects.
Investigations show that the creation of modern competitive technical vision systems for precision tra- jectory measurements is only possible by using simple and reliable algorithms, focused on the use of suffi- ciently high quality and cheap video recorders and hardware and software for optimal price performance, functionality and speed in dealing with this class of problems. In practice, for the creation of such projects, now they usually focus on the use of Raspberry Pi 3 single-board computers and the Python programming language with the Open Source Computer Vision Library. This makes it possible to solve the assigned tasks of determining the location with the necessary accuracy in real time.
The purpose of this work is to study the noise immunity and speed of the algorithms of trajectory meas- urements based on the use of various methods for determining the location of the object of observation in the frame. The task of trajectory measurements is reduced to determining the location of an object on a sequence of frames, as well as its displacements on adjacent frames of the video sequence.
