Remote Sensing and Social Sensing Data Fusion for Fine-Resolution Population Mapping With a Multimodel Neural Network

Abstract

Mapping population distribution at fine spatial scales is significant and fundamental to solve resource utilization, assessment of city disaster, environmental regulation, and urbanization. Multi-source data produced by remote and social sensing have been widely used to disaggregate census information to map population distributions at fine resolution. However, it is challenging to achieve accurate high spatial resolution population mapping by combining multi-source data and considering geographic spatial heterogeneity. The existing approaches do not consider global and local spatial information simultaneously, resulting in low accuracy. This paper proposes a multi-model fusion neural network for estimating fine-resolution population estimates from multi-source data. Our approach takes into account the local spatial information and global information of each geographic unit. Specifically, a first-order space matrix of a geographic unit is used to characterize its local spatial information. We proposed a multi-model neural network, which combines a convolutional neural network (CNN) and a multilayer perceptron (MLP) model to estimate a fine-resolution population mapping. Using Shenzhen, China, as the experimental setting, a population distribution map was generated at 100m spatial resolution. The model was quantitatively validated by showing that it captured the relationship between estimated population and census population at township level (R 2 = 0.77) more accurately than the WorldPop dataset (R 2 = 0.51) and the MLPbased model (R 2 = 0.63). Qualitatively, the proposed model can identify differences in population density in densely populated areas and some remote population clusters more accurately than the WorldPop populcation dataset.

Fractal Characteristic Analysis of Urban Land-Cover Spatial Patterns with Spatiotemporal Remote Sensing Images in Shenzhen City (1988–2015)

“…Urban spatial structure can be considered the external manifestation of the spatial structure formed by various spatial activities under different factors. As urban spatial pattern changes, a set of indicators, including population, infrastructure, and economy, also change [11][12][13]. The resulting scale law can interpret the inherent laws of the urban structure change.…”

From auxiliary data to research prospects, a review of gridded population mapping

“…In addition to improving a single model, the performance of models can also be improved by combining multiple models. For example, the combination of MLP (Multilayer Perceptron) with CNN and MLR with CNN considers the global and local features of the population distribution, which can obtain more accurate results than the single model (Cheng et al, 2021; Li et al, 2020). We believe this improved modeling approach will be dominated by deep learning algorithms, especially CNN algorithms.…”

Soil classification with hyper-temporal satellite images

“…Third, considering the coarse temporal resolution of Landsat images (i.e., ~16-day), the use of available Landsat images easily suffers limitation. To address this point, an advanced spatiotemporal fusion method is effective to simulate sufficient Landsat images [10], by fusing the satellite images with high temporal resolution. Moreover, the classification accuracy has the potential to be further increased.…”