基于优化SIFT算法的无人机遥感作物影像拼接

发布时间：2018-05-28 15:36

本文选题：作物 + 遥感　；参考：《农业工程学报》2017年10期

【摘要】：针对作物遥感影像因对比度低所导致的使用尺度不变特征变换算法(scale-invariant feature transform,SIFT)提取特征点数目少,拼接效果不理想的情况,提出了一种基于图像锐化的自适应修改采样步长的非极小值抑制拼接算法,该算法在图像预处理中引入锐化滤波器对平滑后的图像进行卷积,增强图像细节,增加特征点提取数目,同时通过基于尺度的自适应修改采样步长,使图像特征点分布更加均匀,根据低对比度作物遥感影像的成像特性,采用非极小值抑制,提高图像匹配效率。在查找匹配点的过程中,引入最优节点优先算法(best-bin-first,BBF)查找最近邻与次近邻,采用随机抽样一致算法(random sample consensus,RANSAC)优选特征点。通过试验验证,该文改进后的算法相比于标准SIFT算法,在处理低空作物遥感影像时,特征点提取数目平均增加77.5%,特征点匹配对数平均增加15对,对于标准SIFT算法无法匹配的低对比度作物遥感影像,提取到了8对以上的匹配点对,满足了拼接条件。该改进算法相对于标准SIFT算法更适于低对比度遥感影像的拼接。
[Abstract]:Because of the low contrast of crop remote sensing image, the scale-invariant feature transform (sift) algorithm is used to extract the feature points and the splicing effect is not satisfactory. In this paper, an adaptive modified sampling step size suppression algorithm based on image sharpening is proposed. In this algorithm, sharpening filter is introduced into the image preprocessing to convolution the smooth image to enhance the image details. The number of feature points is increased, and the sampling step is modified adaptively based on scale to make the distribution of feature points more uniform. According to the imaging characteristics of low contrast crop remote sensing images, the non-minimum value is used to suppress the feature points. Improve the efficiency of image matching. In the process of finding matching points, the optimal node-first algorithm is introduced to find the nearest neighbor and the next nearest neighbor, and the random sample consensus algorithm is used to select the feature points. The experimental results show that compared with the standard SIFT algorithm, the improved algorithm increases the number of feature points and the logarithm of feature points by an average of 77.5 and 15 pairs respectively in processing low-altitude crop remote sensing images. For the low contrast crop remote sensing images which can not be matched by the standard SIFT algorithm, 8 pairs of matching points are extracted, which satisfy the stitching condition. Compared with the standard SIFT algorithm, the improved algorithm is more suitable for low contrast remote sensing image stitching.
【作者单位】：东北农业大学电气与信息学院;
【基金】：国家重点研发计划专项(2016YFD0200701);国家重点研发计划专项(2016YFD020060305) 863计划项目(2013AA102303)
【分类号】：TP751

【相似文献】