一种改进KNN的无人机图像快速拼接方法

为了更好解决基于K近邻算法特征匹配速度问题，采用图像像素点经纬度数据加快特征点匹配的无人机图像拼接方法。利用拍摄图片信息里的地理坐标，计算影像像素点经纬度数据，然后计算出两张图像重合部分，利用重合部分特征点经纬度数据大致相同这一特点提高K近邻算法匹配速度，改进后的算法在匹配准确度比传统算法提高了43%左右，最后选用最佳缝合线法对图像进行拼接，获得了质量较好的全景图。

Improved Stitching Method Based on KNN for UAV Images

To improve the efficiency of feature matching based on the K-nearest neighbor （KNN） algorithm， an unmanned aerial vehicle （UAV） image stitching method by using the latitude and longitude information image pixels was proposed. Firstly， the latitude and longitude information of image pixels was calculated based on the geographic coordinates in the captured UAV images. Then the overlapped parts of a pair of images were computed. The matching speed of KNN algorithm can be improved because the longitude and latitude data of overlapped feature points are almost the same. Compared with the traditional KNN algorithm， the matching accuracy increases by about 43%. Finally， the best suture algorithm is selected for image stitching， and high-quality panorama images are obtained.