Segmentation and Classification of Solid Construction Waste with mmWave Imaging

Solid construction waste is imposing a significant weight on both society and the natural environment. For solid waste segmentation and classification, the preferred method is optical imaging. However, the dust and rain can cause attenuation and obscure light for optical imaging. This paper primarily delves into the resource recycling for solid waste using mmWave imaging segmentation methods which weather and dust have little effect on. We propose an intelligent identification method for solid construction waste based on millimeter wave SAR imaging and an improved U-Net network. Our experiments show that mmWave imaging can be used in obstruct cases for solid waste imaging. With improved U-Net and data enhancement methods, we efficiently tackled the solid waste identification issue in construction and achieved over 84% segmentation accuracy on average for 5 classes.