Orientation of vehicles using a camera

Abstract

The aim was to find suitable algorithms that help to find a vehicle coordinates based on the optical camera images. It is assumed that the vehicle firstly drives through a determined route creating a sequence of optical camera images that will be complemented by relevant coordinates. At repeated drives-through, the system already compares the optical camera images with those saved at the first drive-through. Based on this comparison, current vehicle coordinates can be determined. The task appeared to be much more demanding then it looked in the beginning. This problem has probably not been dealt with, therefore it was impossible to find the required support. The originally expected openCV programme libraries assistance has not been helpful. For this reason, it was necessary to solve numerous other problems to be able to complete the original task. However, new algorithms were found, that have not been solved so far. The set goal has been reached. The algorithms derived in this paper were launched into operation at a small vehicle model (using a children building box) with a web camera (Herkules de Luxe). Regarding extensive cost of a bit-maps processor, the vehicle can only move quite slowly to enable the computer to process the camera images timely. Speed increase is only a matter of the processor equipment, either by its speed increase or by a parallel run of more processors. The later solution is feasible, which is referred to in the paper. The application was program med in C++ language. The Borland C++ Builder compiler was used for this purpose. Standard libraries of the compiler are already equipped with programmes for JPEG images conversion into bit-maps.