Back-end optimization method based on multi-modal feature association

The invention discloses a back-end optimization method based on multi-modal feature correlation, which can increase the accuracy and robustness of positioning in a synchronous positioning and mapping system using multi-modal features, and the idea of the back-end optimization method is to construct a correlation model of the features, combine the correlation relationship with back-end optimization, reduce the accumulated error of the system, and perform pose optimization. According to the method, firstly, a mathematical model of a spatial geometrical relationship between features is given, then, feature residual terms corresponding to different features are calculated, and finally, the confidence weight of the error terms is adjusted by using the feature correlation degree in a rear end and pose optimization is performed. According to the method, the re-projection error containing the incidence relation is added into back-end optimization, so that the accuracy and robustness of SLAM system pose calculation ca