图像块改进Harris角点检测的农田图像拼接

Farmland image mosaic based on Harris corner detection improved by image blocks

为了达到农田研究人员对于宽视野、高清晰度图像的要求,提出一种基于图像块改进Harris角点检测的农田图像拼接算法。为了降低图像拼接数据量,减少运算时间,提高拼接效率,将参考图像与待拼接图像分割成具有重叠区域的图像块,采用余弦相似度,寻找最优匹配图像块;采用Harris角点检测算法实现最优匹配图像块的特征点提取,通过计算两者之间的欧氏距离实现特征点粗匹配;采用随机采样一致算法(RANSAC)去除误匹配点,对实现精确匹配,并求解参考图像块与待配准图像块的投影变换矩阵;将投影变换矩阵应用于原始参考图像与待拼接图像,实现拼接。实验结果表明,该研究拼接结果的标准差、平均梯度和空间频率值都高于传统Harris算法;在算法性能中,平均误配率为39.56%和平均运行时间为2.46 s,都优于Harris算法,说明该研究拼接图像的质量较好且算法能够较好地实现农田图像的拼接。

A farmland image mosaic algorithm based on Harris corner detection improved by image blocks is proposed to meet the requirements of farmland researchers for wide field of vision and high resolution images.In order to reduce the amount of image mosaic data and operation time,and improve the efficiency of image mosaic,the reference images and the images under mosaic are segmented into the image blocks with overlapping regions,and the best matching image block is searched by means of the cosine similarity.The Harris corner detection algorithm is used to extract the feature points from the best matching image blocks,and calculate the Euclidean distance between feature points to realize the coarse matching of them.The random sampling consistency algorithm(RANSAC)is used to remove the mismatching points to realize the accurate matching,and solve the projection transformation matrix of the reference image blocks and the image blocks under matching.The projection transformation matrix is applied to the original reference images and the images under mosaic to realize the image mosaic.The experimental results show that the standard deviation,average gradient and spatial frequency of the mosaic algorithm proposed in this paper are all higher than those of the traditional Harris algorithm,and the average mismatching rate is 39.56%and running time is 2.46 s,both of which are better than those of Harris algorithm.It indicates that the quality of the mosaic images in this paper is better and the algorithm can realize the farmland images mosaic better.