基于间隙度的无人机林地航拍图像序列拼接方法

无人机林地航拍图像具有的分辨率高、数据量大、边缘丰富的特点，造成了特征点配准中误匹配率的增加，因此本文提出了一种新的无人机林地航拍图像序列拼接方法。分形中的间隙度可用于描述图像区域纹理的粗糙程度，本文首先利用间隙度特征选取图像中局部图像块作为候选区域查找特征点，减少了待配准的特征点数量，提高了特征点配准正确率；其次，采用全局拼接技术变换图像，减少传统拼接中矩阵连乘产生的误差的积累和传播。实验中选取了不同拍摄高度的两组图像序列，将本方法与传统的全局SURF特征方法和降采样图像拼接方法进行了对比，结果显示本方法可以有效拼接图像，同时不会损失原图像的精度信息，并从视觉效果和均方根误差两个角度证明了本文方法优于其他两种方法。 

A method for woodland aerial image sequences mosaic based on lacunarity

Woodland aerial images captured by UAV( unmanned aerial vehicle) possess high resolution, large amount of pictures, plenty of edges. These features increased the error rate of image registration. In this paper, a new image stitching method based on lacunarity in fractal geometry was proposed. Lacunarity can describe the roughness of texture. We calculated lacunarity value from blocks in image and found several candidate blocks, which were used to extracte the feature points instead of the whole image. This method made contributions to reducing matching feature points amount and improving the accuracy when registration. We used global mosaic method instead of traditional method in the step of image transformation to decrease error accumulation spreading caused by transformation matrix multiplication. The experimental results showed that, when dealing with high resolution UAV image shot from different heights, compared with traditional SURF and down-sample stitching method, the proposed method gave a promising and an excellent visual effects result without losing resolution, and the value of RMSE was better.