Automatic Engineering Quantity Calculation and Rapid Cost Estimation Based on UAV Oblique Photogrammetry and Deep Learning for 3D Point Clouds
Point clouds obtained from UAV oblique photogrammetry have several problems in the “calculable- audiable - rapidly estimated” link, specifically inconsistencies in scale, occlusion and voids leading to increased errors, and a lack of interpretable evidence for the conversion of engineering quantities into costs. Based on this, this paper presents an integrated data-driven workflow “from image to cost”. Fieldwork requires the use of the target GSD and a combined vertical and oblique perspective to ensure geometric closure, and the input criteria are determined based on GCP/CP residuals and effective coverage. Indoor work involves engineering preprocessing, namely denoising and ground separation, to unify the scale of raster and voxels, perform block segmentation, and ensure continuity, thus transforming the reconstructed point cloud into a stable input form. The identification stage forms a measurable entity system based on the bill of quantities standards, utilizing a multi-scale 3D backbone and lightweight image-3D fusion, along with boundary enhancement and rule verification, to finally output verifiable entities. Verification shows that the error between physical quality and quantity of work can be achieved through table closure. For example, the relative error of Cut/Fill is about ±1%. The cost estimation uses bill of quantities mapping and quantile residual band to complete the consistency assessment, and will further provide risk stratification and availability threshold under degradation conditions.