İç mekân harita yapımı için makine öğrenmesiyle nokta bulutlarının sınıflandırılması

Yıl 2024, Cilt: 11 Sayı: 1, 30 - 40, 03.05.2024

Sena Varbil , Alper Şen

https://doi.org/10.9733/JGG.2024R0003.T

Öz

İç mekânlara ait 3-Boyutlu nokta bulutu sınıflandırması, iç mekân harita yapımı, iç mekân navigasyonu, bina yenileme, tesis yönetimi vb. uygulamalarda iç mekân modellerinin oluşturulmasında büyük önem taşımaktadır. Bu çalışmada, Stanford Üniversitesi tarafından üretilen S3DIS (Stanford 3D Indoor Scene) veri setinde bulunan ofis odalarına ait nokta bulutları makine öğrenmesi yöntemlerinden Rasgele Orman (RO) ve Çok Katmanlı Algılayıcı (ÇKA) ile sınıflandırılarak iç mekân haritaları oluşturulmuştur. Giriş verileri için X, Y, Z ve R, G, B öznitelik bilgileri kullanılmıştır. Sınıflar tavan, zemin, duvar, kapı, pencere, kolon, masa, sandalye, pano, dağınıklık ve kitaplık nesnelerini kapsamaktadır. Eğitim ve test verilerinde iç mekân haritalarının oluşturulması amacıyla duvar, kapı, pencere, kolon, pano ve kitaplık bir sınıf (birleştirilmiş sınıf-1); masa, sandalye ve dağınıklık bir sınıf (birleştirilmiş sınıf-2) halinde birleştirilmiştir. Eğitim verisi için bir ofis kullanılmış ve beş ayrı ofiste test edilmiştir. RO yöntemiyle ortalama %88, ÇKA yöntemiyle ortalama %85 sınıflandırma doğruluğu elde edilmiştir. Böylece özellikle yüksek doğrulukla sınıflandırılan tavan ve birleştirilmiş sınıf-1 nesneleri sayesinde iç mekân haritaları da yüksek doğrulukla elde edilmiştir.

Anahtar Kelimeler

İç mekân, Nokta bulutu, Sınıflandırma, Makine öğrenmesi, Rastgele Orman, Çok Katmanlı Algılayıcı

Classification of indoor point clouds using machine learning for indoor mapping

Öz

3-Dimensional point cloud classification of interior spaces is of great importance in the creation of interior models in applications such as indoor mapping, indoor navigation, building renovation, facility management, etc. In this study, point clouds of office rooms in the S3DIS (Stanford 3D Indoor Scene) dataset produced by Stanford University were classified with Random Forest (RF) and Multilayer Perceptron (MLP) machine learning methods, and indoor maps were created. For input data, attributes X, Y, Z and R, G, B were used. The classes include ceiling, floor, wall, door, window, column, table, chair, board, clutter, and bookcase objects. To create indoor maps in the training and test data, the classes were merged as follows: wall, door, window, column, board, and bookcase were merged into one class (merged class-1), and table, chair, and clutter were merged into another class (merged class-2). An office was used for the training data and tested in five different offices. The RF method achieved an average classification accuracy of 88%, and the MLP method achieved an average accuracy of 85%. Thus, indoor maps were obtained with high accuracy, especially thanks to the ceiling and merged class-1, which were classified with high accuracy.

Anahtar Kelimeler

Indoor, Point cloud, Classification, Machine learning, Random Forest, Multi-Layer Perceptron

Kaynakça

• Armeni, I., Sener, O., Zamir, A. R., Jiang, H., Brilakis, I., Fischer, M., & Savarese, S. (2016). 3d semantic parsing of large-scale indoor spaces. Proceedings of the IEEE conference on computer vision and pattern recognition, 1534-1543.
• Bilgili, A., Şen, A., & Başaraner, M. (2022). İç mekân navigasyonu ağ modelleri: Karşılaştırmalı bir inceleme. Jeodezi ve Jeoinformasyon Dergisi, 9(2), 108-126.
• Bishop, C. M. (1995). Neural Networks for Pattern Recognition. Oxford University Press.
• Breiman, L. (2001). Random forests. Machine learning, 45, 5-32.
• Chen, J., & Clarke, K. C. (2020). Indoor cartography. Cartography and Geographic Information Science, 47(2), 95-109.
• Chen, X. T., Li, Y., Fan, J. H., & Wang, R. (2021). RGAM: A novel network architecture for 3D point cloud semantic segmentation in indoor scenes. Information Sciences, 571, 87-103.
• Deng, Y., Ai, H., Deng, Z., Gao, W., & Shang, J. (2022). An overview of indoor positioning and mapping technology standards. Standards, 2(2), 157-183.
• Díaz-Vilariño, L., Khoshelham, K., Martínez-Sánchez, J., & Arias, P. (2015). 3D modeling of building indoor spaces and closed doors from imagery and point clouds. Sensors, 15(2), 3491-3512.
• Fawagreh, K., Gaber, M. M., & Elyan, E. (2014). Random forests: from early developments to recent advancements. Systems Science & Control Engineering: An Open Access Journal, 2(1), 602-609.
• Gunduz, M., Isikdag, U., & Basaraner, M. (2016). A review of recent research in indoor modelling & mapping. The International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences, 41, 289-294.
• Hastie, T., Tibshirani, R., Friedman, J. H., & Friedman, J. H. (2009). The elements of statistical learning: data mining, inference, and prediction. New York: Springer.
• Hsieh, C. S., & Ruan, X. J. (2023). Automated Semantic Segmentation of Indoor Point Clouds from Close-Range Images with Three-Dimensional Deep Learning. Buildings, 13(2), 468.
• Kiran, A., & Vasumathi, D. (2020). Data mining: min–max normalization based data perturbation technique for privacy preservation. Proceedings of the Third International Conference on Computational Intelligence and Informatics: ICCII 2018, 723-734, Singapore: Springer Singapore.
• Lin, H., Wu, S., Chen, Y., Li, W., Luo, Z., Guo, Y., Wang, C., & Li, J. (2021). Semantic segmentation of 3D indoor LiDAR point clouds through feature pyramid architecture search. ISPRS Journal of Photogrammetry and Remote Sensing, 177, 279-290.
• Otero, R., Lagüela, S., Garrido, I., & Arias, P. (2020). Mobile indoor mapping technologies: A review. Automation in Construction, 120, 103399.
• Poux, F., Mattes, C., & Kobbelt, L. (2020). Unsupervised segmentation of indoor 3D point cloud: Application to object-based classification. The International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences, 44, 111-118.
• Sen, A., & Gumus, K. (2023). Comparison of Different Parameters of Feedforward Backpropagation Neural Networks in DEM Height Estimation for Different Terrain Types and Point Distributions. Systems, 11(5), 261.
• Shan, D., Zhang, Y., Wang, X., Luo, W., Meng, X., Liu, Y., & Gao, X. (2023). An Efficient Point Cloud Semantic Segmentation Method Based on Bilateral Enhancement and Random Sampling. Electronics, 12(24), 4927.
• Su, Y., Jiang, L., & Cao, J. (2022). Point Cloud Semantic Segmentation using Multi Scale Sparse Convolution Neural Network. arXiv:2205.01550.
• Sun, Z., Wang, G., Li, P., Wang, H., Zhang, M., & Liang, X. (2024). An improved random forest based on the classification accuracy and correlation measurement of decision trees. Expert Systems with Applications, 237, 121549.
• Taud, H., & Mas, J. F. (2018). Multilayer perceptron (MLP). Geomatic approaches for modeling land change scenarios, 451-455.
• Wu, H., Yang, H., Huang, S., Zeng, D., Liu, C., Zhang, H., Guo, C., & Chen, L. (2020). Classification of point clouds for indoor components using few labeled samples. Remote Sensing, 12(14), 2181.
• Zhao, J., Zhang, X., & Wang, Y. (2020). Indoor 3D point clouds semantic segmentation bases on modified pointnet network. The International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences, 43, 369-373.