Su Lijuan, Xu Yue, Yuan Yan, Yang Jingyi
Key Laboratory of Precision Opto-mechatronics Technology Sponsored by Ministry of Education, School of Instrumentation Science and Opto-electronics Engineering, Beihang University, Beijing 100191, China.
Sensors (Basel). 2020 Mar 9;20(5):1503. doi: 10.3390/s20051503.
Mixed pixels commonly exist in low-resolution remote sensing images, and they are the key factors hindering the classification of land covers and high-precision mapping. To obtain the spatial information at the subpixel level, subpixel mapping (SPM) technologies, including the pixel-swapping algorithm (PSA), that use the unmixed proportions of various land covers and allocate subpixel land covers have been proposed. However, the PSA often falls into a local optimum solution. In this paper, we propose a SPM method, the PSA_MSA algorithm, that combines the PSA and the modified simulated annealing algorithm to find the global optimum solution. The modified simulated annealing algorithm swaps subpixels within a certain range to escape the local optimum solution. The method also optimizes all the mixed pixels in a randomized sequence to further improve the mapping accuracy. The experimental results demonstrate that the proposed PSA_MSA algorithm outperforms the existing PSA-based algorithms for SPM. The images with different spatial dependences are tested and the results show that the proposed algorithm is more suitable for images with high spatial autocorrelation. In addition, the effect of proportion error is analyzed by adding it in the experiments. The result shows that a higher proportion error rate leads to larger degradation of the subpixel mapping accuracy. Finally, the performance of PSA_MSA algorithm with different ranges of selection on subpixel-swapping is analyzed.
混合像素普遍存在于低分辨率遥感影像中，是阻碍土地覆盖分类和高精度制图的关键因素。为了获取亚像素级别的空间信息，人们提出了亚像素映射（SPM）技术，包括利用各种土地覆盖的解混比例并分配亚像素土地覆盖的像素交换算法（PSA）。然而，PSA常常陷入局部最优解。在本文中，我们提出了一种SPM方法，即PSA_MSA算法，它将PSA与改进的模拟退火算法相结合以找到全局最优解。改进的模拟退火算法在一定范围内交换亚像素以逃离局部最优解。该方法还以随机顺序优化所有混合像素，以进一步提高映射精度。实验结果表明，所提出的PSA_MSA算法优于现有的基于PSA的SPM算法。对具有不同空间依赖性的图像进行了测试，结果表明所提出的算法更适合具有高空间自相关性的图像。此外，通过在实验中添加比例误差来分析其影响。结果表明，较高的比例误差率会导致亚像素映射精度的更大程度下降。最后，分析了PSA_MSA算法在不同亚像素交换选择范围内的性能。
