CONSTRUCTION PROGRESS MEASUREMENT USING DEEP LEARNING
Abstract
Failing to keep track of the construction progress in time and making current construction progress falling behind the schedule will cause a significant loss of time and money. Monitoring the progress of the constructing structure in construction sites is one of the significant challenges today. The work of comparing buildings in construction with drawings requires time-consuming and expensive manual inspections. Computer vision techniques for construction progress monitoring were developed to solve these problems. By using the photographic apparatus, on-site situations can be detected and monitored remotely.
However, the complexity of the picture makes it difficult to quantify the progress of the project. Semantic segmentation, as a branch of deep learning, is a method to identify objects in images at the pixel level. With the semantic segmentation technology of computer vision, the scene in the picture can be simplified, and the objects in the picture can be finely segmented. In this paper, to quantify the amount of work in the picture taken in construction sites, a new test was performed. This test is to estimate the completion rate of specific construction structures by comparing the segmented construction structures in the pictures with the BIM model. The objective of the research is to evaluate the rationality of the test.
Subject
Construction Progress MonitoringConstruction Completion Rate
Computer Vision
Deep Learning
Semantic Segmentation
AI
Citation
Mou, Xue (2019). CONSTRUCTION PROGRESS MEASUREMENT USING DEEP LEARNING. Master's thesis, Texas A&M University. Available electronically from https : / /hdl .handle .net /1969 .1 /189128.