Xuan Zhou

To extend the scale limits within which a template watermark remains detectable, this paper proposes a template watermarking algorithm for remote sensing images based on a radial spectral barcode (RSB). The RSB is a new geometry of template watermark whose inner and outer boundaries are separated along the radial direction of the discrete Fourier transform magnitude spectrum. In the embedding stage, the spectrum is divided into angular slots, a synchronization sequence generated from the secret key and the watermark are assigned to alternate slots, and the magnitude within the radial interval of each slot is modulated according to its sign. In the detection stage, the angular profile of the received spectrum is computed under a set of geometric hypotheses, the synchronization sequence is located by circular correlation, and the watermark is read from the synchronized profile without correcting the image. The experimental results demonstrate that the proposed method extends the detectable scale interval well beyond that of the ring-template methods while retaining robustness against rotation, translation, cropping, affine transformation, map reprojection, and signal-processing attacks. It outperforms the comparative algorithms in robustness at an equal or higher PSNR, with a low false detection rate.

Global environmental change has shifted disaster risk from being predominantly natural hazard-dominated to increasingly socially driven. While the Social Vulnerability Index (SoVI) is widely used to quantify these dynamics, its construction remains hindered by uncertainty in weighting algorithms and spatial aggregation schemes. This study assesses SoVI in Taiwan across three harmonized study years (2015, 2020, and 2025) by bridging data innovation, spatial representation, and methodological synthesis. To offer fine-scale spatial representation, we integrate gridded nighttime light, urban built-up volume, and population density to shift vulnerability assessments from conventional administrative borders to functional, data-driven Human Agglomeration Zones (HAZs). Furthermore, we examined algorithmic uncertainty by comparing the variance-driven Principal Component Analysis (PCA) against the Entropy Weight Method (EWM). We found that integrating gridded socioeconomic data can effectively identify small contiguous HAZ patches that were commonly omitted by administrative divisions. The results also reveal a twofold pattern of vulnerability. PCA produced comparatively stable rankings and repeatedly identified south-central counties as high-vulnerability areas, whereas EWM generated more concentrated weights and a stronger reordering of vulnerability patterns in 2025 as the weight assigned to population density increased sharply. Ultimately, by proposing a synthesized multi-method framework, a comprehensive SoVI was established to mitigate algorithmic bias, providing policymakers with a multidimensional tool that aligns long-term structural investments with short-term dynamic monitoring.