Urban functional zone mapping by integrating high spatial resolution nighttime light and daytime multi-view imagery

Abstract

Functional zones are the basic units of cities and the mapping work is fundamental to urban management, investigation, and research. The existing urban functional zone (UFZ) mapping methods usually utilize visual features from high spatial resolution optical images, and focus on two-dimensional image features, such as texture and landscape. However, UFZ is a comprehensive concept including geographical, social, and economic aspects. Therefore, it is reasonable to simultaneously take into account the characteristics of both human activities and image visual features for its accurate interpretation. Multi-view optical satellite images can delineate the physical characteristics of a city, i.e., 3D structures; on the other hand, high spatial resolution nighttime light images are important signals of human activities. These two data sources can be complementary in representing urban landscapes. However, to our knowledge, in the current literature, neither multi-view images nor high spatial resolution nighttime light images have been used for UFZ mapping, and it is poorly understood whether their individual or combined use can achieve satisfactory results. Therefore, in this study, a daytime and nighttime data fusion method, that is, the fusion of daytime multi-view optical images (Ziyuan3-01, 2.1 m) and high spatial resolution nighttime light images (Jilin1-07, 0.92 m), was proposed for UFZ mapping. In particular, a building enhanced nighttime light index (BENI) was proposed to improve the ability of the nighttime light images in discriminating between different functional zones. To verify the effectiveness of the proposed method, experiments were conducted in two megacities of China, i.e., Beijing and Wuhan. Our results indicated that: 1) an OA of ~80% was obtained by the spectral based method; 2) the addition of multi-view features led to ~84% OA, an increment of 4% compared with the spectral features; and 3) the inclusion of nighttime features achieved 85–90% OA, which further improved the OA of daytime features by 1–6%. It was also shown that the proposed BENI was superior to the original nighttime light brightness in identifying functional zones. In general, this study verified the effectiveness and complementarity of daytime (including multispectral and multi-view images) and nighttime images in UFZ mapping, and provided new thoughts for day and night data fusion and urban mapping.