Akademik Veri Yönetim Sistemi
41st International Symposium on Automation and Robotics in Construction, ISARC 2024, Lille, Fransa, 3 - 05 Haziran 2024, ss.759-767
This study introduces an innovative method for enhancing digital modeling accuracy in construction site monitoring by integrating UAV imaging with advanced machine learning and computer vision algorithms. It focuses on removing temporary elements like construction machinery from images. The method involves two steps: first, using deep learning algorithms, for instance, segmentation to detect and segment construction machinery from UAV images trained on the Aerial Image Dataset for Construction (AIDCON); second, employing image inpainting techniques, utilizing the Places2 dataset and the LaMa algorithm, to fill in the areas left vacant by the removed machinery. Demonstrated on a parking garage construction site in Ankara, Türkiye, the results show high accuracy in machinery segmentation and effective image inpainting, as evidenced by metrics like Normalized Root Mean Square Error (NRMSE), Peak Signal-to-Noise Ratio (PSNR) and Structural Similarity Index (SSIM). This approach contributes significantly to the field of construction site monitoring by refining digital models and shows potential for broader application in the industry. Future research directions include developing a specialized image inpainting dataset for construction scenarios and extending the methodology to encompass more types of temporary site elements, paving the way for more efficient and accurate project management in construction.