Automatic identification of illegal construction and demolition waste landfills: A computer vision approach

Dozens of landslide accidents are reported at construction and demolition waste (CDW) landfills worldwide every year. Those accidents could be avoided via timely inspection in which the identification of illegal CDW landfills at a large scale plays a critical role. Traditional field surveys are time-consuming, labor-intensive, which is not effective in large-scale detection of landfills. To address this issue, a methodology is proposed in this study for the automatic identification of CDW landfills in large-scale areas by utilizing semantic segmentation of remote sensing imagery. Deep learning is employed to achieve automatic identification and a case study is conducted to showcase the models. The results shown that: (1) The model proposed in this study can effectively identify CDW landfills, with an accuracy of 96.30 % and an IoU of 74.60 %. (2) DeepLabV3+ demonstrated superior performance over Pspnet and HRNet, though HRNet approached DeepLabV3+ in performance with appropriate optimizations. (3) Case study results indicate the potential existence of 52 CDW landfills in Shenzhen, includng 4 official landfills and 48 suspected illegal CDW landfills, mainly in Longhua, Guangming, and Baoan districts. The method proposed in this study provides an effective approache to identify large-scale illegal CDW landfills and has great significance for supervising CDW landfills.

Automated summary

This study proposes a methodology for the automatic identification of construction and demolition waste (CDW) landfills in large-scale areas using semantic segmentation of remote sensing imagery. The use of deep learning achieves automatic identification, and the case study results demonstrate the effectiveness of the method in identifying CDW landfills.