Integrasi Deep Learning Mask R-CNN dalam Pemodelan 3D LOD1 Berbasis Citra Foto Udara UAV

Mohammad Misbahuddin; Septa Erik Prabawa; Slamet Riadi; Yahya Alfrid Koroh; Shilvy Choiriyatun Navisa

doi:10.24014/jej.v5i2.37883

Integrasi Deep Learning Mask R-CNN dalam Pemodelan 3D LOD1 Berbasis Citra Foto Udara UAV

Mohammad Misbahuddin, Septa Erik Prabawa, Slamet Riadi, Yahya Alfrid Koroh, Shilvy Choiriyatun Navisa

Abstract

Pemodelan 3D Level of Detail 1 (LOD1) merupakan representasi geometris dasar berupa bentuk prismatik dengan atap datar, yang umum digunakan dalam perencanaan wilayah dan manajemen aset karena efisiensi komputasi serta kesederhanaan strukturnya. Namun, proses pembentukan model ini kerap menghadapi kendala dalam aspek efisiensi waktu, akurasi spasial, serta skalabilitas, khususnya pada tahap segmentasi dan identifikasi objek dari data citra 2D. Penelitian ini bertujuan untuk mengimplementasikan algoritma deep learning Mask R-CNN dalam rangka meningkatkan efisiensi dan akurasi proses ekstraksi jejak bangunan (building footprint) dari citra udara beresolusi tinggi yang dihasilkan oleh wahana UAV. Mask R-CNN memiliki keunggulan dalam mendeteksi objek serta menghasilkan segmentasi berbasis pixel (pixel-wise mask), sehingga memungkinkan delineasi batas objek secara presisi. Alur kerja penelitian meliputi akuisisi data UAV, segmentasi citra secara otomatis menggunakan Mask R-CNN, dan rekonstruksi 3D berbasis Sistem Informasi Geografis (SIG). Studi kasus dilakukan di Kelurahan Jawa, Kota Samarinda, dengan resolusi spasial citra sebesar 0,023 meter, akurasi horizontal (CE90) 0,139 meter, dan vertikal (LE90) 0,282 meter. Tiga skenario pelatihan menunjukkan tingkat deteksi bangunan masing-masing sebesar 33%, 42%, dan 55% dengan nilai presisi 0,589; 0,746; dan 0,794. Skenario ketiga (80% data pelatihan) menghasilkan visualisasi model 3D paling mendekati hasil digitasi manual. Namun, skenario kedua (70% data pelatihan) direkomendasikan karena waktu pemrosesan tercepat (2 jam 27 menit 58 detik) dan kebutuhan penyimpanan terkecil (1,82 GB). Secara ekonomi, metode ini mengurangi biaya sebesar 13% dibandingkan digitasi manual, dengan total biaya Rp13.367.504.

Keywords

Deep Learning; LOD1; Mask R-CNN; Pemodelan 3D, UAV.

Full Text:

PDF

References

Abdul Rahman, R., Mohammad Yusof, Y., Kashefi, H., & Baharum, S. (2012). Developing Abdulla, W. (2017). Mask R-CNN for object detection and instance segmentation on Keras and TensorFlow.

Anurogo, W., Lubis, M. Z., Khoirunnisa, H., Hanafi, D. S. P. A., Rizki, F., Surya, G., & Dewanti, N. A. (2017). A simple aerial photogrammetric mapping system overview and image acquisition using unmanned aerial vehicles (UAVs). Geospatial Information, 1(1), 11-18.

Danielczuk, M., Matl, M., Gupta, S., Li, A., Lee, A., Mahler, J., & Goldberg, K. (2019, May). Segmenting unknown 3d objects from real depth images using mask r-cnn trained on synthetic data. In 2019 International Conference on Robotics and Automation (ICRA) (pp. 7283-7290). IEEE.

Deliry, S. I., & Avdan, U. (2021). Accuracy of unmanned aerial systems photogrammetry and structure frommotion in surveying and mapping: a review. Journal of the Indian Society of Remote Sensing, 49(8), 1997-2017.

Eltner, A., & Sofia, G. (2020). Structure from motion photogrammetric technique. In Developments in Earth surface processes (Vol. 23, pp. 1-24). Elsevier.

He, K., Gkioxari, G., Dollár, P., & Girshick, R. (2018). Mask R-CNN. IEEE Transactions on Pattern Analysis and Machine Intelligence, 42(2), 386–397. https://doi.org/10.1109/TPAMI.2018.2844175

Huang, G., Liu, Z., Van Der Maaten, L., et al.: ‘Densely connected convolutional networks’. IEEE Proc. Conf. Computer Vision and Pattern Recognition, Honolulu, HI, USA, 7 July 2017, vol. 1, p. 3

Javadnejad, F. (2017). Small Unmanned Aircraft Systems (UAS) for Engineering Inspections and Geospatial Mapping. Dissertation of Oregon State University, 1-132.

Jiang, S., Jiang, W., & Wang, L. (2021). Unmanned Aerial Vehicle-Based Photogrammetric 3D Mapping: A survey of techniques, applications, and challenges. IEEE Geoscience and Remote Sensing Magazine, 10(2), 135-171.

Khayyal, H. K., Zeidan, Z. M., & Beshr, A. A. (2022). Creation and spatial analysis of 3D city modeling based on GIS data. Civil Engineering Journal, 8(1), 105.

Kraff, N. J., Wurm, M., & Taubenbock, H. (2020). Uncertainties of Human Perception in Visual Image Interpretation in Complex Urban Environments. IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing, 13, 4229–4241. https://doi.org/10.1109/jstars.2020.3011543

Le, T. D., Huynh, D. T., & Pham, H. V. (2018, November). Efficient human-robot interaction using deep learning with mask R-CNN: Detection, recognition, tracking and segmentation. In 2018 15th International conference on control, automation, robotics and vision (ICARCV) (pp. 162-167). IEEE.

Nex, F., & Remondino, F. (2014). UAV for 3D mapping applications: a review. Applied geomatics, 6, 1-15.

Nys, G. A., Billen, R., & Poux, F. (2020). Automatic 3d buildings compact reconstruction from LiDAR point clouds. The International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences, 43, 473-478.

Peters, R., Dukai, B., Vitalis, S., van Liempt, J., & Stoter, J. (2022). Automated 3D reconstruction of LoD2 and LoD1 models for all 10 million buildings of the Netherlands. Photogrammetric Engineering & Remote Sensing, 88(3), 165-170.

Remondino, F. (2011). Heritage recording and 3D modeling with photogrammetry and 3D scanning. Remote sensing, 3(6), 1104-1138.

Templin, T. (2023). LOD 0 LOD1. The Routledge Handbook of Geospatial Technologies and Society, 348.

Ullman, S. (1979). The interpretation of structure from motion. Proceedings of the Royal Society of London. Series B. Biological Sciences, 203(1153), 405-426.

Xavier, A. I., Villavicencio, C., Macrohon, J. J., Jeng, J. H., & Hsieh, J. G. (2022). Object detection via gradient-based mask R-CNN using machine learning algorithms. Machines, 10(5), 340.

Zhang, C., Zhou, J., Wang, H., Tan, T., Cui, M., Huang, Z., ... & Zhang, L. (2022). Multi-species individual tree segmentation and identification based on improved mask R-CNN and UAV imagery in mixed forests. Remote Sensing, 14(4), 874.

DOI: http://dx.doi.org/10.24014/jej.v5i2.37883