Research On Environment Modeling And Navagation Of Home Robot Based On Monocular Vision

With the development and maturity of robot technologies, the applications of robot systems have been extended to families and the service field. Research on the technologies of home robot navigation has become a hot field in the past few years. Autonomous navigation is very important for mobile robots and is the kernel technique for robotic intelligence. These technologies mainly include environment information collection, environment information description, environment information modeling and automatic positioning, among which the first and second item are the key items.Researches on the autonomous navigation and environmental modeling of monocular visual home mobile robot based on indoor environments, a mixed environment model based on feature-topology was proposed in this paper.In this paper, two environment describing methods are proposed for the mixed environment model of feature-topology: environment description based on the vertical edge feature and the invariance feature.The environment description based on the vertical edge feature adopts the Sobel operator and non-maxima suppression algorithm to extract and refine of the edge, complete the edge of the vertical projection of objects and uses world coordinates of the vertical edge feature to describe the environment and a proportion invariance matching method of the projection curve is proposed.The SURF feature is used to describe the environment based on the local gray distribution feature of the environment, the text introduce the SURF feature's extraction in detail, the descriptor's structure, the feature matching, the structure of the natural landmark and experimental comparison and analysis with SIFT.Finally, this paper gives a home robot navigation system architecture and analysis of the role of each module. This system use the mixed environment model for robot self-positioning and HMM estimation for node identification and the vertical edge for the inter-node location. The experiment proves that the method can improve the positioning accuracy to ensure the continuity of the entire positioning process and increase robustness and real-time performance.