An Optimization Method for SLAM Feature Point Extraction in Fuzzy Scenes

In order to solve the problem that image blur is easily caused by camera shaking of micro-robots during surgical navigation, which leads to the accumulation of SLAM errors, this paper studies the feature point extraction optimization method based on deep learning in fuzzy scenes, proposes an improved defuzzification network, and finally applies the network to the front end of SLAM, and conducts experimental verification on the open data set TUM. The results show that the proposed algorithm significantly improves the accuracy of front-end feature extraction and matching.