结合Retinex增强的井下图像拼接方法

Underground mine image mosaic technique in combination with Retinex

为在矿井环境下尽量多的提取图像特征点数量,从而监控矿井下生产情况.采用局部双边滤波算法对图像进行增强,再利用近似的Hessian矩阵和框状滤波确定特征点的位置;计算特征点的描述子向量,采用最近距离比次近距离的匹配算法将特征点配对,使用RANSAC算法消除误匹配错误;利用特征点计算出变换矩阵,采用线性渐变融合方法进行图像融合.研究结果表明:图像增强后特征点数量明显增多,SURF算法的拼接效率显著上升,有利于提高匹配的准确性和拼接的快速性.

In order to extract the feature points under the environment of the mine as many as possible for monitoring production of underground mining,this paper adopted the partial bilateral filtering algorithm to enhance images,and utilized the approximate Hessian matrix and frame-like filtering to determine the positions of the feature points,then calculated the descriptor vectors of feature points,employed matching algorithm of the ratio of the closest distance and the next closest distance to match feature points,and eliminated error matching by RANSAC algorithm.Finally the paper made use of the feature points to calculate the transformation matrix,and used the method of linear gradient fusion to realize image fusion.The research results show that,the number of the feature points significantly increase after enhancing image,and the splicing of SURF algorithm efficiency is also increased noticeably,which are helpful to improve the accuracy of matching and the quickness of splicing.