Neural Radiance Field (NeRF) is a groundbreaking paradigm in neural implicit representations that revolutionized 3D reconstruction, rendering, and dynamic scene modeling. To address cross-domain fragmentation and unclear technical pathways, we present a systematic framework that surveys theoretical foundations, benchmark datasets, methodological advances, and application scenarios. We begin by analyzing NeRF's core mechanisms, including radiance field modeling and differentiable volume rendering, and by defining standardized evaluation benchmarks. Then we chart evolutionary pathways in model optimization, input adaptation, and dynamic scene modeling and analyze how key methods are linked. Furthermore, we provide task-specific insights that highlight migration bottlenecks and potential remedies across digital content creation, embodied perception, and other specialized domains. We also provide comprehensive references and forward-looking guidance for further theoretical refinements and cross-disciplinary deployment of NeRF-based technologies.
Yao et al. (Thu,) studied this question.