Abstract: With the increasing demand for green energy, the photovoltaic panel infrastructure in the highway area has become an important way to develop renewable energy. As an important part of the highway, the photovoltaic power generation of toll stations and service areas has also been paid attention to. This paper studies the technical method of using deep learning method to identify the information of photovoltaic panels at tollbooths and service areas in the highway road domain through high-resolution remote sensing images. Taking Jiangsu province as the research experimental area, Google 19 remote sensing image data of the whole province are downloaded. By making samples, the existing classical semantic segmentation networks HRNet, ResNet, FCN and U-Net are used to extract information from the experimental area, and the photovoltaic panel information extraction results are obtained. Ablation experiments confirm that the HRNet semantic segmentation network combined with CBAM attention mechanism proposed in this paper has the best extraction effect. This method provides technical support for the intelligent monitoring and management of photovoltaic panels in toll stations and service areas of expressways.

Key words: photovoltaic panels in highway, high-resolution remote sensing images, improved HRNet semantic segmentation network, CBAM attention mechanism, Jiangsu province experimental area