基于Gabor变换与克隆选择的高保真遥感影像阴影角度精确校正方法

Accurate Shadow Angle Correction of High Fidelity Remote Sensing Image based on Gabor Transform and Clone Selection

面对遥感影像受到噪点影响无法有效补偿阴影角度,使得校正结果陷入局部最优而导致角度校正耗时长的问题,提出了基于Gabor变换与克隆选择的高保真遥感影像阴影角度精确校正方法。采用Gabor滤波函数提取影像纹理特征,根据局部化特点计算空间纵横比,将影像纹理集中在中频区域,推导出Gabor滤波中心频率。通过选择Gabor滤波参数,完成对遥感卫星信号噪声滤除处理。结合自适应全局阈值法分割遥感影像序列,获取前景和背景区域两组序列,计算两组序列方差,分割前景和背景区域。引入阴影色彩索引,计算索引梯度值,滤除噪点影像阴影区域,保持影像原有轮廓特征。构建阴影估计模型,对滤波处理影像阴影区域补偿,完成阴影区域角度校正。利用克隆选择算法的选择、克隆、变异操作,优化校正结果,避免陷入局部最优。由试验结果可知,该方法阴影区域无噪点,校正后影像与初始影像一致,且耗费最短时间为5 s,具有精确校正效果。

In view of the problem that remote sensing images cannot effectively compensate the shadow Angle due to the influence of noise,which makes the correction result fall into the local optimal and leads to the long time of Angle correction,an accurate correction method of shadow Angle of remote sensing images based on Gabor transform and clone selection was proposed.The Gabor filter function was used to extract image texture features,the spatial aspect ratio was calculated according to the localization characteristics,the image texture was concentrated in the middle frequency region,and the Gabor filter center frequency was derived.By selecting Gabor filtering parameters,the remote sensing satellite signal noise filtering is completed.Combined with the adaptive global threshold method,remote sensing image sequences were segmented to obtain foreground and background sequences,calculate the variance of the two sequences,and segment the foreground and background regions.The shadow color index is introduced,the index gradient value is calculated,the shadow area of the noise point image is filtered,and the original contour features of the image are maintained.The shadow estimation model was constructed to compensate the shadow region of the image processed by filtering,and the Angle correction of the shadow region was completed.The selection,cloning and mutation operations of the clone selection algorithm are used to optimize the correction results and avoid falling into the local optimal.According to the experimental results,the shadow area of this method has no noise,the corrected image is consistent with the initial image,and the shortest time is 5s,so it has accurate correction effect.