Semi-dense Map Reconstruction of Bronchus Based on Prior Feature Correlation

Simultaneous localization and mapping (SLAM) technology can perform map visualization and bronchoscopy positioning based on monocular bronchoscopy images, reduce the complexity of doctor ’s operation, and improve the success rate and safety of surgery. Aiming at the problems of sparse mapping and uncertain scale caused by complex bronchial anatomical structure and weak intracavity texture, this paper proposes an improved monocular bronchoscopy SLAM algorithm. Inspired by the idea of Kalman filtering, a search and matching strategy combining prior feature correlation and local binary pattern (LBP) is proposed to inverse the depth of some pixels, which improves the density of mapping under the premise of ensuring accuracy. The dynamic experimental platform built by the simulated bronchial model verifies the feasibility of the method.