Construction Waste Image Classification Algorithm Based on Fusion of Multiple Features
Yi You, Boon Cheong Chew · 2025
The rapid development of global urbanization has triggered the rapid expansion of the construction industry. The consequent construction waste problem has caused serious pollution to the environment and occupied a large amount of valuable land resources. In this context, effective classification and treatment of construction waste has become particularly urgent, which can help reduce environmental burdens and promote sustainable economic development. This study proposes a fine-grained construction waste classification algorithm based on image processing technology. The algorithm achieves automatic classification through a multi-feature fusion model and can accurately distinguish five major types of construction waste, including concrete, bricks, metal, wood, and plastic. This study further explores the impact of spatial positional relationships, semantic correlation, and internal structure analysis on improving classification accuracy. These research results not only provide technical support for the effective processing of construction waste, but also provide new methods and perspectives for other fine-grained image classification tasks. Experimental results prove that the algorithm proposed in this study significantly improves the classification accuracy. Specifically, the algorithm achieved high accuracy of 98.97%, 98.69%, 98.70%, 98.72% and 98.68% respectively in the classification tasks of concrete, bricks, metal, wood and plastic, which is significantly better than those based on CNN and enhanced FCM. Although these models are close to the performance of the algorithm in this study in some cases, the overall accuracy is lower.