Robust Point Cloud Registration for Aircraft Engine Pipeline Systems.

Area of Science:

• Aerospace Engineering
• Computer Vision
• Robotics

Background:

• Aircraft engine systems rely on complex pipelines requiring regular inspection for safety.
• 3D scanning captures pipeline data but faces challenges like occlusions and similar structures in point clouds.
• Accurate registration of multi-view point clouds is essential for complete pipeline representation.

Purpose of the Study:

• To develop a novel registration framework for aircraft engine pipeline point clouds.
• To address challenges posed by occlusions and geometric similarities in pipeline structures.
• To enhance the accuracy and reliability of aircraft pipeline inspection.

Main Methods:

• Extraction of pipeline axis point feature structure using cylindrical characteristics.
• Development of a new 3D descriptor, Point Line-Point Pair Features (PL-PPFs).
• Integration of pipeline and engine assembly line features into the PL-PPFs descriptor.

Main Results:

• The proposed framework effectively registers incomplete and occluded aircraft pipeline point clouds.
• The PL-PPFs descriptor accurately identifies the structure of the engine's piping system.
• Experimental results validate the approach's effectiveness on real-world aircraft engine pipeline data.

Conclusions:

• The novel registration framework significantly improves the processing of aircraft pipeline point clouds.
• The PL-PPFs descriptor offers a robust solution for feature extraction in complex, occluded scenes.
• This work contributes to safer and more efficient aircraft maintenance through enhanced inspection capabilities.