基于尺度不变特征变换的宽幅遥感影像云阴影检测方法

摘要:

遥感影像中云阴影的尺度因拍摄高度、云层厚度等因素变化,使得在不同拍摄角度下,不能转却提取特征点,而及时适应云阴影的尺度变化,导致检测不能能准确检测云阴影。为此,提出基于尺度不变特征变换的检测方法。首先,采用SIFT算法和梯度直方图对云阴影进行特征点提取；其次,根据提取的特征点,计算每个特征点的SIFT描述子,并对参考图像中的云阴影特征点的SIFT描述子进行标准化处理；然后；计算这些标准化后的SIFT描述子之间的间距；最后,采用云阴影区域灰度补偿的方法,平衡不同光线条件下的影像质量,并引入均方误差对经灰度补偿后的结果进行检验,完成云阴影的检测。实验结果显示,本文设计方法在光照充足条件下能够精准检测出两个云阴影位置,光线较暗条件下检测结果与实际仅存在最大为1px的误差,能够精准定位两个云阴影的位置,从而提高了云阴影检测准确性。

Abstract:

The scale of cloud shadows in remote sensing images varies due to factors such as shooting height and cloud thickness, making it difficult to extract feature points at different shooting angles and adapt to the scale changes of cloud shadows in a timely manner, resulting in inaccurate detection of cloud shadows. Therefore, a detection method based on scale invariant feature transformation is proposed. Firstly, SIFT algorithm and gradient histogram are used to extract feature points from cloud shadows; Secondly, based on the extracted feature points, calculate the SIFT descriptors for each feature point, and standardize the SIFT descriptors for cloud shadow feature points in the reference image; then; Calculate the spacing between these standardized SIFT descriptors; Finally, the method of cloud shadow area grayscale compensation is adopted to balance the image quality under different lighting conditions, and the mean square error is introduced to verify the results after grayscale compensation, completing the detection of cloud shadows. The experimental results show that the method designed in this paper can accurately detect the positions of two cloud shadows under sufficient lighting conditions. Under darker lighting conditions, there is only a maximum error of 1px between the detection results and the actual situation, which can accurately locate the positions of the two cloud shadows and improve the accuracy of cloud shadow detection.