Automatic Segmentation of 3D Point Clouds of Rubble Masonry Walls, and its Application to Building Surveying, Repair and Maintenance

Changing climatic conditions are contributing to faster deterioration of building fabric. Increasing number of heavy rainfall events can particularly affect historic and Cultural Heritage (CH) buildings. These evolving and uncertain circumstances demand more frequent survey of building fabric to ensure satisfactory repair and maintenance. However, traditional fabric surveys have been shown to lack efficiency, accuracy and objectivity, hindering essential repair operations. The recent development of reality capture technologies, together with the development of algorithms to effectively process the acquired data, offers the promise of transformation of surveying methods.

This paper presents an original algorithm for automatic segmentation of individual masonry units and mortar regions in digitised rubble stone constructions, using geometrical and colour data acquired by Terrestrial Laser Scanning (TLS) devices. The algorithm is based on the 2D Continuous Wavelet Transform (CWT), and uniquely it does not require the wall to be flat or plumb. This characteristic is important because historic structures, in particular, commonly present non-negligible levels of bow, waviness and out-of-verticality.

The method is validated through experiments undertaken using data from two relevant and highly significant Scottish CH buildings. The value of such segmentation to building surveying and maintenance regimes is also further demonstrated with application in automated and accurate measurement of mortar recess and pinning. Overall, the results demonstrate the value of the automatic segmentation of masonry units towards more comprehensive and accurate surveys.