遥感影像缺失重建方法对比与时间序列合成

The Methods Comparison of Remote Sensing Image Missing Reconstruction and Time Series Image Synthesis

遥感影像利用率低的重要原因之一是在卫星的成像过程中,传感器故障或者厚云等天气原因经常造成缺失的情况,降低了影像的利用率。对比分析三种方法(直接修复、局部线性直方图、加权线性回归方法)的重建效果,并针对以上方法仅利用单幅影像重建,没有综合利用多时相信息、无法同时实现去云和条带的不足,提出了利用时间序列影像生成合成影像的方法,达到去云和条带的效果。结果显示,对比方法中,加权线性回归方法修复效果最佳,其重建结果与真值相关系数最高达0.87;所提方法可同时去除云和条带区域,相关系数最高达0.76,因此对于改造时间序列影像数量和质量具有较好作用。

One of the important reasons for the low utilization rate of remote sensing images is that in the imaging process of satellites,weather conditions such as sensor failure or thick clouds often cause missing situations,reducing the utilization of images.In this paper,the reconstruction effects of three methods(direct repair,local linear histogram,weighted linear regression method)were compared and analyzed.For the above method,only a single image is used for reconstruction,and the multi-temporal information is not comprehensively utilized and the missing reconstruction of cloud and strip can not be realized at the same time.So a method for generating synthetic images with time-series images was proposed to achieve the effects of removing clouds and stripes.The results showed that the weighted linear regression method had the best repair effect among the compared methods,and the correlation coefficient between the reconstruction result and the true value was up to 0.87;the proposed method could remove cloud and strip regions simultaneously while the correlation coefficient reached 0.76 which proved that the synthetic image had a good effect on transforming the quantity and quality of time-series images.