In recent years, Neural Radiance Fields (NeRF) have garnered widespread attention in the domains of deep learning and 3D scene rendering, marking a significant advancement in computer vision and graphics. Given the increasing prominence of NeRF and the surge in related research, a thorough summary and analysis of recent progress are warranted. In this paper, we provide a comprehensive examination of the technological evolution, key challenges, and recent developments associated with NeRF. We begin with an introduction to the theoretical foundations of NeRF, followed by an overview of its various adaptations addressing different application scenarios, such as virtual reality, augmented reality, filmmaking, and game development, as well as the practical limitations it faces, including high computational demands and slow rendering speeds. Subsequently, we review publicly available datasets pertinent to NeRF methods and summarize the characteristics of different variants based on experimental results. Finally, we highlight potential research directions and future developments, indicating the extensive application potential of optimized NeRF and its variants across diverse fields. Our objective is to provide both academia and industry with a detailed understanding of the future research and application prospects of NeRF in graphics and scene analysis.