摘要
针对现有图像能见度测量方法存在目标物架设复杂和物理量解算困难的问题,该文提出一种简单有效的基于图像特征的大气能见度估算方法。采用图像梯度和对比度作为能见度的特征变量,将图像感兴趣区进行分窗处理后,利用支持向量机(SVM)和随机森林(RF)两种算法建立能见度数值与图像特征的关系模型。结果表明,图像梯度和对比度特征能够准确反映能见度数值大小;窗口总数对模型精度有影响。窗口总数小于35时SVM算法优于RF算法;在窗口总数为70、窗口大小为140×10时,RF算法的像素精度最高,最优模型决定系数R2为0.965,均方根误差为658.13 m。
In view of that the current methods of atmospheric visibility measurement based on the image have many problems,such as the complex erection of the target and difficult calculation of physical quantities,the atmospheric visibility measurement based on image features is proposed here.The gradient and contrast of the image are selected as image features.The support vector machine(SVM)algorithm and the random forest(RF)algorithm are used to build the model of the relationship between visibility values and image features.All results show that the gradient and contrast of the image are highly correlated with the atmospheric visibility value.The total number of sub-windows has effect on the accuracy of the model.The result of the SVM algorithm is typically better than that of the RF algorithm when the total number of sub-windows is less than 35.The sub-window with the size of 140×10 pixels performs best for the random forest algorithm.The R 2 of the optimal model is 0.965 and the root-mean-square error is 658.13 m.
作者
石玉立
王彬
卜帆
Shi Yuli;Wang Bin;Bu Fan(School of Remote Sensing&Geomatics Engineering,Nanjing University of Information Science&Technology,Nanjing 210044,China)
出处
《南京理工大学学报》
EI
CAS
CSCD
北大核心
2018年第5期552-559,共8页
Journal of Nanjing University of Science and Technology
基金
国家自然科学基金(41471312)
关键词
能见度
梯度
对比度
支持向量机
随机森林
visibility
gradient
contrast
support vector machine
random forest