基于Accord.NET框架的全景图自动拼接

Panorama Stitch Based on Accord.NET Framework

基于Accord.NET框架,创建了可以全自动拼接2幅不同景观图片的算法流程,用Harris算法对图片进行特征提取,使用最大关联准则匹配相似像素点,并利用RANSAC算法对伪匹配集提纯,构造鲁棒性优越的单应性矩阵以优化匹配效果,最后利用改进的线性Alpha通道融合完成图片缝合,实现了拼接处的平滑过渡,并将各处理流程封装成Accord.NET处理类,便于2次开发与推广.实验表明,图片合成处理流程效率高,成像效果理想,具有较高的实用价值.

This article creates a completely automatic progress of stitching two different images .It picks the features of the picture by means of Harris Corners Detector w hich involved maximum bidirectional cor‐relation pixels .Besides this ,it will use RANSAC algorithm to filter the correlated points in order to make a robust homography matrix for better matching result .Finally ,one image will be covered partly by an‐other image by means of a linear gradient Alpha blending which can implement smoothly transition chan‐ging between the two pictures .All process will be packaged in Accord .NET class to further development and applications .The experiment shows the highly effective and wonderful creation results .It is an ad‐vanced practical performance to be out .