Abstract
The rapid development of remote sensing techniques provides rich, large-coverage, and high-temporal information of the ground, which can be coupled with the emerging deep learning approaches that enable latent features and hidden geographical patterns to be extracted. This article marks the first attempt to cross-compare performances of popular state-of-the-art deep learning models in estimating population distribution from remote sensing images, investigate the contribution of neighboring effect, and explore the potential systematic population estimation biases. We conduct an end-to-end training of four popular deep learning architectures, i.e., VGG, ResNet, Xception, and DenseNet, by establishing a mapping between Sentinel-2 image patches and their corresponding population count from the LandScan population grid. The results reveal that DenseNet outperforms the other three models, while VGG has the worst performances in all evaluating metrics under all selected neighboring scenarios. As for the neighboring effect, contradicting existing studies, our results suggest that the increase of neighboring sizes leads to reduced population estimation performance, which is found universal for all four selected models in all evaluating metrics. In addition, there exists a notable, universal bias that all selected deep learning models tend to overestimate sparsely populated image patches and underestimate densely populated image patches, regardless of neighboring sizes. The methodological, experimental, and contextual knowledge this article provides is expected to benefit a wide range of future studies that estimate population distribution via remote sensing imagery.
Highlights
F INE knowledge of the spatial contribution of human activity is essential for a wide range of fields, such as public health [1]–[3], urban planning [4]–[6], disaster managementManuscript received March 2, 2021; revised April 4, 2021 and April 25, 2021; accepted April 27, 2021
As for neighboring scenarios, the increase of neighboring sizes leads to reduced population estimation performance, which is found universal for all four selected models in all evaluating metrics
In terms of the neighboring effect, our results suggest that the increase of neighboring sizes leads to reduced population estimation performance, which is found universal for all four selected models, in all evaluating metrics, and in both Metro Atlanta and Metro Dallas, contradicting a recent study by Xing et al [28], who found 3 × 3 neighboring scenario outperformed 1 × 1 via ResNet architecture
Summary
F INE knowledge of the spatial contribution of human activity is essential for a wide range of fields, such as public health [1]–[3], urban planning [4]–[6], disaster managementManuscript received March 2, 2021; revised April 4, 2021 and April 25, 2021; accepted April 27, 2021. Despite the authority of census-based population distribution released by the officials, it owns several intrinsic limitations, making it ill-suitable for many spatial problems. The census-based population suffers from the modifiable areal unit problem (MAUP) [14] due to its arbitrarily imposed boundaries that are rarely consistent with other boundaries in practical applications [15]. Census-based population distribution is often with poor temporal resolutions that preclude temporal-dynamic population estimations, and recent and reliable population data at fine scales can often be lacking, especially in resource-poor settings [16], [17]. Scholars start to explore various means to improve the aggregated census-based population, one notable effort of which is to derive fine-grained, spatially-continuous population grids
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have