Optimization of odometry and place recognition algorithms through analysis of images and 3D point clouds

In the context of 3D sensing, the main objective is to build maps as accurately as possible.To achieve this goal, algorithms must be robust in different environments and application contexts. Therefore, various sensors such as cameras and LiDAR are used to provide different types of data. The integration of these data types helps improve the robustness of odometry algorithms. In this application context, deep learning techniques have become increasingly important due to advances in knowledge and the improvement of the quality and diversity of data provided by sensor manufacturers. The main objective is to leverage these techniques to enhance odometry and place recognition algorithms, strengthening the generated map and enabling real-time navigation within it.