A visual SLAM using graph method

Monocular Simultaneous Localization and Mapping is a key technique in the research field of mobile robots. Based on the natural feature of the environment, this paper proposes a visual SLAM algorithm based on graph. The algorithm extracts the visual features of the collected images through the SURF method. Using the image coordinates of the same feature points in different time, the position relationship of the robot in different time can be attained. As a result, a camera relative pose estimation map can be built using these position constraints. Subsequently, the map can be optimized using the stochastic gradient descent method in order to form a consistently global map. The experimental results demonstrates the validity and practicability of the presented method.