I. M. Lebedev Post-graduate Student, P.G. Demidov Yaroslavl State University A. L. Tyukin Post-graduate Student, P.G. Demidov Yaroslavl State University A. L. Priorov Dr.Sc. (Eng.), Associate Professor, P.G. Demidov Yaroslavl State University

Аt the present time the problem of global navigation has been solved. But we have troubles with it, when it concerns the robots indoor movement. It-s generally known, that work in confined spaces is complicated by a variety of interferences (for example, reflections of radio signal). The analysis of current scientific literature has shown that the most reliable channel of communication is an optical channel. Therefore the navigation system of autonomous mobile platform by the color markers was proposed. With the help of special color markers for indoors navigation, we can find objects (on that they are located), using the computer vision algorithms, and evaluate their position and orientation in space. Analysis of the system in real conditions was carried out and the functionality borders for the algorithm were also defined. The analysis shows, that the algorithm is sensitive to luminosity and to the ambient lighting. It was ascertained, that the computer vision algorithm can identify markers correctly only with illumination above 22 lux. Preferred light sources are: incandescent lamp; natural light without direct sunlight. Dependence of markers recognition on the distance "camera - marker" was also investigated. This distance is very important, because when it increases, the marker-s relative area, it takes in the image decreases. And the findings show, that at the distance more than 1.5 m dispersion increases sharply and at the distance more than 1.7 m algorithm is not able to recognize the marker. Consequently, the height of the marker should be defined on the basis of characteristic distances used in the task. Obstacle detection is an important task for mobile robots. Many robots rely on range-sensor, but all these sensors have their own shortcomings. Therefore we have developed and researched a new obstacle detection system based on color information with using of only one camera. The main point of our obstacle detection algorithm is the research of color composition of ground surface with the following estimation for belonging to it of each pixel on the image. This estimation is performed with the help of calculation of Mahalanobis distance between pixels coordinates in appropriate color space and the model of the ground. Two modifications of our algorithm have been developed. One of them uses RGB color space, and the other uses HSV color space. Dispersion of quantity of pixels that were detected as ground was chosen as quality criterion of algorithm. Experimental research of algorithm shows that the less these value the more reliable obstacle detection. Algorithm was researched by next parameters: Dependence on Mahalanobis distance; Dependence on input image contrast; Dependence on input image noise level. The system was tested with the help of static and dynamic images (video). The research has shown that at PSNR > 38 dB HSV is preferable to RGB, at PSNR < 38 dB RGB is preferable HSV, minimum of dispersion we can get if the image contrast is 60, and also there is such mahalanobis distance at which the algorithm works optimally. Obstacle detection algorithmalso properly works in outdoor.

 

1. SHapiro L., Stokman Dzh. Kompjuternoe zrenie. M.: BINOM. Laboratorija znanijj. 2006.
2. Gonsales R., Vuds R. Cifrovaja obrabotka izobrazhenijj. M.: Tekhnosfera. 2005.
3. Seljaev A.A., Alpatov B.A. Algoritm ocenki mestopolozhenija obekta na dvumernom izobrazhenii // Izv. vuzov. Ser. «Priborostroenie». 1988. № 5. C. 3-5.
4. Alpatov B.A., Seljaev A.A. Stepashkin A.I. Cifrovaja obrabotka izobrazhenijj v zadache otslezhivanija dvizhushhegosja obekta // Izv. vuzov. Ser. Priborostroenie. 1985. № 2, S. 39-43.
5. Nilsson N.J. Shakey the Robot. TechnicalNote. 1984.
6. Everett H.R. Sensors for Mobile Robots: Theory and Applications. Massachusetts. 1995.
7. Alpatov B.A. Ocenivanie parametrov dvizhushhegosja obekta v posledovatelnosti izmenjajushhikhsja dvumernykh izobrazhenijj // Avtometrija. 1991. № 3.S. 21-24.
8. Faugeras O., Robert L., Laveau S., Csurka G., Zeller C., Gauclin C., Zoghlami. 3-D reconstruction of urban scenes from image sequences // Comput. Vision and Image Understanding. 1998. P. 292-309.
9. Bakut P.A., Kolmogorov G.S., Vornovickijj I.EH. Segmentacija izobrazhenijj: Metody porogovojj obrabotki // Zarubezhnaja radioehlektronika. 1987. № 10. S. 6-24.