Figure 9.

Integrated architecture combining the developed LiDAR and vision based pipelines. The initial detection of the pipe is provided by the LiDAR RANSAC segmentation, providing very robust detection with average accuracy positioning. Using data from the FMU odometry or any other state estimation process available (e.g., any SLAM approach, visual odometry, etc.) the estimated pose of the pipe is used as a prior in the visual pipeline, simplifying the visual detection problem. Thus, the greater accuracy of the visual pose recovery can be exploited.