基于井下环境的SIFT算法研究

Research of SIFT Algorithm Based on Underground Environment

针对煤矿环境中现有图像特征匹配算法不适用的问题,提出了新的特征匹配算法。该算法首先对图像进行Curvelet去噪预处理,然后采用SIFT进行特征向量的构建和匹配,通过RANSAC方法的优化改进去除误匹配点,通过建立投影变换模型实现图像的拼接。实验表明,该算法在噪声大、光照不均、模糊的复杂环境中有较好的鲁棒性,解决了煤矿环境中图像容易误匹配的现象,拼接效果平滑自然。

The paper put forward a new feature matching algorithm to solve the problem that existing feature matching methods are not suitable to underground environment. Firstly, the algorithm uses Curvelet algorithm to preprocess image noise, then adopts SIFT algorithm to form and match feature vectors, which removes mismatched points effectively by optimization of RANSAC method and achieves image stitching by building projection transformation model. The experiment showed that the algorithm has better robustness in noise, poor illumination and fuzzy environment, and solves mismatching phenomenon in mine environment, whose stitching result is smoothing and natural.