Visual SLAM optimization method and system based on spatial multivariate geometric features

The invention provides a visual SLAM method and system based on spatial multivariate geometric features, and belongs to the technical field of computer vision, and the method comprises the steps: constructing a unified parameterized expression model for three different types of feature point clouds of points, lines and surfaces, and in the front-end matching tracking link, a unified residual model is designed according to different types of feature matching relationships to carry out fusion registration. And a local optimization equation is defined according to the unified residual error model in a rear-end local optimization link, and joint optimization of three types of feature elements of points, lines and surfaces in a local map under the unified optimization equation is realized. According to the method, feature unification is carried out on extracted feature point clouds of different types of points, lines and surfaces through a unified parameterized model, a unified residual model is proposed for all t