关键词: 特征场景, 主成分分析, 全局特征, SURF, 匹配范围划分

Abstract: In this paper, a new method based on eigenscene for fast image matching was presented, and it could solve the problem that mainstream matching algorithm with local features cannot describe the global features. By extracting scene images, the eigenscene using Principal Component Analysis (PCA) could be reconstructed, and then it could be used for matching scope division. In the divided matching scope, SURF algorithm was used to match fast local features. The experimental results indicate this method combines the large scale global features and scale invariant local features so that the discrimination ability is enhanced for similar object. The robustness and timeliness of this method achieve good balance so that it extends the application fields of mainstream matching method with local features. In the end, improvement direction is suggested for this paper, which shows the scalability of this method.

Key words: eigen-scene, Principal Component Analysis (PCA), global feature, Speeded-Up Robust Features (SURF), mathing scope division