| APPLICATION OF MACHINE AND DEEP LEARNING STRATEGIES FOR THE CLASSIFICATION OF HERITAGE POINT CLOUDS |
| Paper ID : 1210-SMPR |
| Authors: |
|
Fabio Remondino *, Eleonora Grilli, Emre Ozdemir FBK Trento, Italy |
| Abstract: |
| Nowadays the use of cultural heritage point clouds for documentation, communication and valorisation purposes is increasing. The association of heterogeneous information to 3D data by means of automated classification methods can help to characterize, describe and better interpret the object under study. Great progress has been brought to classification procedures with machine learning methods. Different methods were proposed like edge and region-based approaches or model fitting approaches, based on the possibility to fit geometric primitives to the 3D shapes. However, in the field of cultural heritage this blooming of approaches has not been fully explored. This paper presents an ongoing research for the automated classification of various heritage point clouds with different machine and deep learning approaches (Random Forest, OvO, RNN Bi-LSTM, 1D and 2D CNN). The aim is to automatically recognize architectural components such as pavements, columns, facades, windows, etc. and to evaluate the performaces of these methods. Qualitative and quantitative results are reported for two complex scenarios. |
| Keywords: |
| Point Clouds, Classification, Machine Learning, Deep Learning, Cultural Heritage |
| Status : Conditional Accept (Oral Presentation) |
