CBGW-YOLOv8: An Improved Model for Intelligent Monitoring of Construction Projects in Remote Sensing Images

Effective project monitoring has become crucial as the construction industry expands and the number of projects increases. Current monitoring methods, primarily relying on manual labor and traditional remote sensing techniques, face challenges such as insufficient image resolution and low processing speed. This study addresses these issues by applying deep convolutional neural networks to high-resolution remote sensing image recognition and developing an intelligent monitoring model based on YOLOv8 (CBGW-YOLOv8) for identifying various construction project stages. A dataset named “Construction Site,” containing different construction stages, was created using pre-processed remote sensing data from Jilin-1. An improved YOLOv8 architecture was adopted, with modifications to the network structure and loss function to enhance learning capability for construction features. The model's effectiveness in improving monitoring efficiency and accuracy was validated through actual deployment. Experimental results demonstrate that the CBGW-YOLOv8 model outperforms traditional methods and the original model in construction project detection, achieving a 2.9% increase in mAP<sub xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">0.5</sub> and a 2.6% increase in mAP<sub xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">0.5:0.95</sub>, while exhibiting higher accuracy and faster processing speed.