Development of a system for assessing the quality of urban street-level greenery using street view images and deep learning

Street greenery has long played a vital role in the quality of urban landscapes and is closely related to people's physical and mental health. Also, the level of street greenery is an important indicator of urban environmental quality. However, despite extensive research into environmental assessment methods for urban greenery, plant identification and greenery index calculations are still mostly done manually. In this research, we developed a method based on semantic segmentation processing of street view images to calculate the Green View Index of urban streets, and the Panoramic View Green View Index (PVGVI) is proposed for measuring the visible street-level greenery. We validated the results by comparison with those of manual inspection and the Pyramid Scene Parsing Network method. The vegetation detection rate of our method is very close to the ground truth value, which means it can distinguish almost all of the vegetation information from the street view images, and based on it we can calculate the PVGVI which is reliable. In addition, we conducted a case study of street-level greenery using the PVGVI and confirmed that this method can better visualize urban street-level greenery. The proposed method is scalable and automatable, and it contributes to the growing trend of integrating large freely available street view image datasets with semantic segmentation to inform urban planners.