Algorithm Of Indoor Navigation For Mobile Robot Based On Binocular Vision

With the rapid development of computer science and Automation and Control Engineering, etc, the mobile robot is more and more widely used in various fields of social life. At the same time, the machine vision and pattern recognition technology continues to advance, and vision sensor has become a hotspot of research on mobile robot. Robot navigation and positioning research based on vision sensor is a new topic related to a variety of technologies, such as image processing, stereo matching, map building and path planning.The paper’s main research content is the application of binocular vision sensor in mobile robot navigation and positioning, According to the research status at home and abroad, the paper analyzes the robot obstacle detection, image matching, map building and path planning problems, and puts forward improvement or innovation solution algorithms.Obstacle detection in robot navigation and positioning system is the premise and key factor of the robot motion control, the accuracy and stability of obstacle detection directly affects the precision and stability of the whole system. Aiming at the problem of serious environmental light interfering obstacles extraction, the paper proposes the Retinex enhanced secondary obstacles extraction algorithm under repair. Firstly do Retinex image enhancement processing to obtain the environmental image, which eliminates the influence of uneven illumination, then use the seed filling algorithm to fill the road, background and obstacles in the image, finally do image binarization, edge detection and contour extraction to obtain the shape information and location information of the obstacle. In the corridor environment, this paper adopts the combination of Retinex algorithm and Otsu threshold segmentation algorithm to achieve the extraction of stable ladder crural line information, and completes the full extraction of the corridor road on this basis.Retinex algorithm is often used in the image enhancement processing of remote sensing image and blurred image in the fog or rain environment. This paper adopts the Retinex algorithm to eliminate illumination shadow in the image, and obtains good results. Comparing with the other shadow elimination methods, such as homomorphic filter, illumination compensation, this paper adopts single scale Retinex algorithm to enhance image, and does some experiments to explore the scale parameter of the gaussian template. This paper adopts the combination of linear transformation and gray stretch, and significantly reduces the gray level difference of the environmental pavement point, provides a good basis for the next step of image segmentation.In view of the long time-consuming problem and the large range of image scanning in the process of image matching, this paper calculates and summers up the minimum scanning range based on polar constraint principle and the biggest parallax scope of targets in the area of robot interest, and shortens the single match time to millisecond, significantly improves the system instantaneity.This paper uses the combination of topology and geometry map to describe the environment map, and combines the global path planning and the topological map creation by the Freud algorithm. At the same time, the global topological map is divided into local section between the path nodes to complete the connection of topology and geometry map. In obstacle avoidance process of partial section, this article divides the geometric map obstacles into several layers, and designs the strain type obstacle avoidance strategy to improve the strain ability of robot navigation.In robot localization aspect, this paper designs a geometry artificial landmarks of red, white and black color, and achieves the goal of fast and exact robot positioning by detecting color, shape ratio and intensity feature,. At the same time, the artificial landmarks are the mark of robot navigation in local zoning. Through a simple and efficient artificial sign design, this paper implements the purposes of the robot accurate positioning and transformation of mixed map, and etc.