Three-dimensional grid map building based on binocular vision

A new method of three-dimensional grid map building is brought out in this article. SIFT matching algorithm is used to pick up feature points. To improve the matching results, errors matching deleting and dense matching method are used, which is helpful for three-dimensional grid map building. Then probability model of voxel is built and parameter optimization of the matching points is adopted to modify the value of probability function. This three-dimensional grid map can give service robots detailed 3D space information in indoor unstructured environment. The experiment result proved the validity and feasibility of the algorithm.