航空遥感影像的阴影处理方法研究

发布时间：2018-05-27 16:04

本文选题：航空遥感 + 阴影检测　；参考：《西安电子科技大学》2014年硕士论文

【摘要】：航空遥感影像具有空间分辨率高、信息量大的特点,被广泛应用于地理信息产业、城市信息化建设、服务与旅游事业等城市经济和社会发展事业中。近年来,遥感影像分辨率随遥感传感技术的提高而呈百倍的增长,这更是突出了影像中阴影的存在。阴影削弱甚至消除了被遮挡区域物体的光学和物理信息,对后续的影像处理,包括目标识别、地物分类等造成消极的影响。因此,面对航空遥感技术产出的海量遥感影像,阴影去除技术成为近年来研究的热点与难点。国内外学者针对航空遥感影像提出的阴影处理方法,能够成功完成阴影检测与补偿。在此基础上,如何提高阴影检测与补偿的准确率与速度、如何扩展其应用范围又是阴影处理的重要研究方向。本文研究了3c通道的不足,对基于3c通道的阴影检测算法的各步骤提出改进方案:第一,3c通道影像的非线性增强处理。为log变换公式设计可选参数扩展像素值的动态变化范围,然后利用阈值法修正增强结果。第二,平滑去噪处理。以邻接像素点与中心点的距离为影响因子设计权重公式,并结合阈值法来修正边缘细节处图像模糊的效应。第三,阴影边缘检测。针对改进的Sobel边缘检测算法,利用高斯分布和抽样方法估计边缘阈值,从而实现阈值的自动选取。实验结果证明,本文算法可有效降低噪声对阴影提取的影响,尤其是在边缘位置,改进后的算法准确率得到提高,适用于不同场景下的航空遥感影像的阴影检测。本文重点分析了现有的基于颜色恒常的阴影补偿算法对航空遥感影像进行阴影补偿时的优缺点,针对校正增益计算的不准确性以及原算法忽略半影区域在阴影补偿所占的重要地位这两个问题,提出基于颜色恒常的分区域阴影补偿算法:第一,本影的补偿。首先,确定同质区,根据阴影区域的大小确定同质区的大小,然后通过“逐层选取”的方法标记出阴影的同质区;其次,分别对阴影区和同质区进行颜色恒常计算得到颜色校正增益;最后,通过颜色恒常公式将本影区域的色彩变换到非阴影光照条件下。第二,半影的补偿。首先,对阴影边缘施行扩展处理,获得半影区域;然后,在半影区分段多项式模型的基础上,设计半影光照补偿公式;最后,根据光照补偿公式对半影区域进行阴影补偿。航空遥感影像因其地物的多样性与复杂性而在图像处理领域拥有其特殊的地位,有针对性地对其进行阴影检测与补偿始终是图像处理的重点与难点。针对阴影检测,本文分析了阴影区域的共有特征,没有研究地物的独有特征,因此,如何有针对性地分析地物的阴影相关特征,从而更有效地检测出阴影也是今后的研究方向。针对阴影补偿,本文使用同质区的颜色恒常计算结果作为非阴影区的结果并增加半影补偿步骤,如何在此基础上进一步提高阴影补偿的精度应进一步研究。
[Abstract]:Aerial remote sensing images are widely used in urban economic and social development, such as geographic information industry, urban information construction, service and tourism, because of their high spatial resolution and large amount of information. In recent years, the resolution of remote sensing image has increased by a hundred times with the improvement of remote sensing technology, which highlights the shadow in the image. Shadow weakens or even eliminates the optical and physical information of the occluded area, which has a negative effect on the subsequent image processing, including target recognition, ground object classification, and so on. Therefore, in the face of massive remote sensing images produced by aerial remote sensing technology, shadow removal technology has become a hot and difficult point in recent years. The shadow processing methods proposed by scholars at home and abroad for aerial remote sensing images can successfully complete shadow detection and compensation. On this basis, how to improve the accuracy and speed of shadow detection and compensation, and how to expand its application range is an important research direction of shadow processing. In this paper, the deficiency of 3c channel is studied, and an improved scheme is proposed for each step of shadow detection algorithm based on 3c channel: first, nonlinear enhancement processing of 3c channel image. For the log transform formula, the optional parameters are designed to extend the dynamic range of pixel values, and then the enhancement results are corrected by the threshold method. Second, smooth denoising. The distance between the adjacent pixel and the center is used as the influence factor to design the weight formula and the threshold method is combined to correct the image blur effect at the edge details. Third, shadow edge detection. For the improved Sobel edge detection algorithm, the Gao Si distribution and sampling methods are used to estimate the edge threshold, thus the automatic selection of the threshold is realized. The experimental results show that the proposed algorithm can effectively reduce the influence of noise on shadow extraction, especially in edge position, and the improved algorithm can be applied to shadow detection of aerial remote sensing images in different scenes. This paper focuses on analyzing the advantages and disadvantages of the existing shadow compensation algorithms based on color constant for aerial remote sensing image shadow compensation. Aiming at the inaccuracy of the correction gain calculation and the fact that the original algorithm neglects the important position of the penumbra region in shadow compensation, a sub-region shadow compensation algorithm based on color constant is proposed. First, the homogenous area is determined, the size of the homogeneous area is determined according to the size of the shadow area, and then the homogeneous area of the shadow is marked by the method of "layer by layer selection"; secondly, The color correction gain is obtained by calculating the color constant of the shadow region and the homogeneous region, respectively. Finally, the color of the shadow region is transformed to the non-shadow illumination condition by the color constant formula. Second, penumbra compensation. First, the shadow edge is extended to obtain the penumbra region; then, based on the polynomial model of the penumbra differentiation segment, the penumbra illumination compensation formula is designed. Finally, the shadow compensation is carried out according to the illumination compensation formula. Aerial remote sensing image has a special position in the field of image processing because of its diversity and complexity. It is always the focus and difficulty of image processing to detect and compensate its shadow pertinently. Aiming at shadow detection, this paper analyzes the common features of shadow area, and does not study the unique features of ground objects. Therefore, how to analyze the shadow correlation features of ground objects in order to detect shadows more effectively is also the research direction in the future. For shadow compensation, we use the result of color constant calculation of homogeneous region as the result of non-shadow region and add penumbra compensation steps. How to further improve the accuracy of shadow compensation on this basis should be further studied.
【学位授予单位】：西安电子科技大学
【学位级别】：硕士
【学位授予年份】：2014
【分类号】：TP751

【参考文献】