Research On Indoor Navigation System Of Vision-Based Mobile Robot

As the development of modern information technology, the robot technology has been rapidly progressed and the application in scientific exploration, wrecker disaster relief, military area, industrial manufacturing and other field were very wide. At the same time, as the people's living standard is improving, the robot technology was more and more applied in entertainment. Vision-based robot navigation technology is a hot spot of modern mobile robot navigation in recent years has been widely applied to robot navigation. The detection of the navigation path was the key of mobile robot navigation methods which directly affected the accuracy of the movement of robot. Because vision based navigation method obtained information integrity and high level intelligence, it certainly broadly development in future robot navigation. As People engaged in several of work, and some of them with heavy, repetitive, unhealthy and even dangerous, and so on, so a device is needed to replace people to engage in such work. The Robot has mobility, flexibility, and self-government capacity, in addition, robot has no idea of tireless and can adapt to the changing environment, so its application is more and more wide. The system was designed for mobile robot exhibition in service industry, and it can work at exhibition halls or museums. The algorithm of the extraction of robot navigation line, exhibition platform mark recognition and audio read and play methods were researched in this paper.This paper discussed the background of robot navigation and historical significance, and also discussed the research situation in details. In addition, this paper profoundly studied the methods of robot navigation, proposed the application condition of various methods. At the same time, it introduced the purpose and significance of this paper.Studied the image processing knowledge and algorithms which commonly used in visual navigation system, according to the characteristics and requirements of this system, this issue introduced the image preprocessing algorithm, de-noising algorithms, segmentation algorithm and color space model. The image in this system were all 24 bits color images, and the de-nosing algorithm used the combination of median filter and mathematical morphology, and compared various of segmentation algorithms and ultimately selected OTSU method for automatic segmentation and Sobel edge detection method.Researched the extraction algorithm of navigation line, and analysis the advantages and disadvantages of tradition Hough transform algorithm and least square method. We proposed a two-centroid method based on path edges according to navigation road environment. The central idea of this method is calculate the midpoints according to segmented edges and handle these points specially and finally calculate the central point of the upper and lower half image, then calculate the navigation line. This method simple and fast,Studied several ways of the mark design and identification, and introduced digital identification, RFID and color recognition methods, and also introduced the advantages and disadvantages of these methods. Thinking the exhibition environment of indoor, we choose color recognition under HSI color space model, not only meet the needs of exhibition, but also the identification of this mark is relatively simple and the algorithm is easy to implement.Researched some of commonly used audio formats and play methods, and then introduced their application area and their advantages and disadvantages. At last, we used MCI method to play WAVE format audio files after clearly thinking of the characteristic of this system and habit of people.This system is a mobile robot visual navigation exhibition system which based on DirectX. The experimental platform was built by set DirectX 9.0 in VC++ 6.0. Thinking the indoor exhibition environment, we segmented out the exhibition platform and road through processing the collecting road images, then identified the mark of exhibition platform and play relative exhibition information to realize exhibition. Finally, we verified the feasibility of this method.