Coastal urbanization is increasingly constrained by legacy land-use patterns and escalating climate risks, yet long-term morphological trajectories remain poorly quantified due to the absence of multispectral data in pre-satellite archives. This study presents a scalable deep learning pipeline that bridges this century-scale domain gap, enabling automated reconstruction of urban expansion from panchromatic historical aerial imagery (1920–1971) and digital aerial photographs (1997) to contemporary very-high-resolution satellite data (2024) in Les Sables-d’Olonne, France. Spectral restoration was performed using an attention-enhanced Pix2Pix generative adversarial network with hybrid inference, achieving high fidelity (PSNR 35.21 dB, SSIM 0.9762). Semantic segmentation was conducted with U-Net++, yielding strong performance on modern data (mIoU 0.9789). However, direct transfer to historical periods suffered from severe domain shift due to radiometric variations.To overcome this limitation without extensive manual annotation, few-shot adaptation was applied on year-specific calibration sets, producing reliable building footprints (mIoU 0.53–0.65) despite degradation. Multi-scalar analysis of the reconstructed footprints revealed constrained anisotropic expansion: early saturation of the coastal historic core, followed by rapid inland peri-urbanization post-1971 driven by geographic barriers. This spatiotemporal shift has entrenched spatial lock-in, placing recent development in retro-littoral zones vulnerable to submersion and characterized by severe vegetation loss. This framework unlocks previously inaccessible historical archives for quantitative urban monitoring, providing critical insights into legacy effects of unconstrained growth and informing resilient coastal planning under climate change.