Feature selection SLAM algorithm incorporating attention and anticipation

Aiming at the problem of localization failure of SLAM in the scene of sharp turning and fast movement, a feature selection simultaneous localization and mapping (SLAM) algorithm incorporating attention and anticipation is proposed. It can select features that are more likely to stay in the field of view as the camera moves, and discard features that are about to disappear from the field of view. The logdet metric is used to measure the feasibility of quantifying feature selection firstly. And then, calculate the information matrix of features. From the detected features, a greedy algorithm is used to select the features (approximately) to maximize the logdet metric. Finally, the actual test combined with ORB-SLAM2 shows that the algorithm can ensure the accuracy of positioning in complex scenarios (such as sharp turning and fast movement).