This survey provides a comprehensive overview of local feature matching methods, categorizing them into Detector-based and Detector-free approaches. Detector-based methods include Detect-then-Describe, Joint Detection and Description, Describe-then-Detect, and Graph-Based techniques, while Detector-free methods encompass CNN-based, Transformer-based, and Patch-based approaches. The paper evaluates existing datasets and metrics to quantitatively compare state-of-the-art techniques and discusses their applications in areas such as Structure from Motion, Remote Sensing Image Registration, and Medical Image Registration. It also identifies current challenges and outlines future research directions, serving as a reference for researchers in local feature matching and related fields. The survey highlights the importance of adapting to diverse and dynamic scenarios and explores the role of large foundation models in feature matching. It emphasizes the ongoing relevance of traditional manual methods alongside deep learning technologies and the emerging focus on multi-modal images. The paper also discusses the evolution of local feature matching methodologies, including the development of robust descriptors, the integration of attention mechanisms, and the use of graph neural networks for matching. It covers various techniques for feature detection and description, including self-supervised learning, weakly supervised learning, and the use of CNNs and Transformers for dense matching. The survey also addresses the challenges of handling extreme viewpoint changes and textureless regions, and explores the potential of deep learning in improving the accuracy and robustness of local feature matching. Overall, the paper aims to provide a thorough analysis of the latest advancements in local feature matching, emphasizing the importance of integrating different approaches and techniques to achieve more effective and efficient feature matching.This survey provides a comprehensive overview of local feature matching methods, categorizing them into Detector-based and Detector-free approaches. Detector-based methods include Detect-then-Describe, Joint Detection and Description, Describe-then-Detect, and Graph-Based techniques, while Detector-free methods encompass CNN-based, Transformer-based, and Patch-based approaches. The paper evaluates existing datasets and metrics to quantitatively compare state-of-the-art techniques and discusses their applications in areas such as Structure from Motion, Remote Sensing Image Registration, and Medical Image Registration. It also identifies current challenges and outlines future research directions, serving as a reference for researchers in local feature matching and related fields. The survey highlights the importance of adapting to diverse and dynamic scenarios and explores the role of large foundation models in feature matching. It emphasizes the ongoing relevance of traditional manual methods alongside deep learning technologies and the emerging focus on multi-modal images. The paper also discusses the evolution of local feature matching methodologies, including the development of robust descriptors, the integration of attention mechanisms, and the use of graph neural networks for matching. It covers various techniques for feature detection and description, including self-supervised learning, weakly supervised learning, and the use of CNNs and Transformers for dense matching. The survey also addresses the challenges of handling extreme viewpoint changes and textureless regions, and explores the potential of deep learning in improving the accuracy and robustness of local feature matching. Overall, the paper aims to provide a thorough analysis of the latest advancements in local feature matching, emphasizing the importance of integrating different approaches and techniques to achieve more effective and efficient feature matching.