iSEAM: Improving the Blooming Effect Adjustment for DMSP-OLS Nighttime Light Images by Considering Spatial Heterogeneity of Blooming Distance
iSEAM: Improving the Blooming Effect Adjustment for DMSP-OLS Nighttime Light Images by Considering Spatial Heterogeneity of Blooming Distance
复制标题
DOI：
10.1109/jstars.2021.3065399
复制
发表时间：
2021
影响因子：
5.5
通讯作者：
L. Zhuo;Chenyang Zhang;Xiaolin Zhu;Tianhao Huang;Yang Hu;Haiyan Tao
中科院分区：
文献类型：
--
作者：
L. Zhuo;Chenyang Zhang;Xiaolin Zhu;Tianhao Huang;Yang Hu;Haiyan Tao
最长的档案使DMSP-OLS夜间灯光(NTL)图像在相关时间序列研究中无与伦比。然而，这些研究一直受到开花效应的限制。2019年提出的自调整模型(SEAM)在一定程度上解决了这一问题。然而，Seam假定NTL图像中的所有像素都是恒定的光晕距离3.5公里。事实上，花开的距离与土地覆盖和人造灯光的亮度有关。这一假设在鲜花距离不同于3.5公里的城市中导致了很大的误差。针对这一问题，本研究通过考虑开花距离的空间异质性，提出了一种改进的接缝模型(ISEAM)。具体地说，iSEAM对DMSP-OLS图像进行分割得到光目标，然后利用随机森林方法估计每个光目标的有效光晕距离，然后通过改进的像素亮度交互模型校正每个光目标中所有像素的光晕效应。在中国的试验表明，中国的开花距离在0到12.55公里之间，平均为3.36公里。经iSEAM校正后的影像与NPP-VIIRS影像的相关系数(R)达到0.70，高于其他光晕效应校正方法。此外，iSEAM校正后的图像具有比其他方法更高的空间异质性。这些结果表明，通过考虑有效光晕距离的空间异质性，iSEAM可以作为一种更准确、更有效的方法来校正DMSP-OLS NTL图像的光晕效应。
The longest archive makes DMSP-OLS nighttime light (NTL) images unparalleled in relevant time series studies. However, these studies have been constrained by the blooming effect. The self-adjusting model (SEAM) proposed in 2019 solves this problem to some extent. However, SEAM assumed all pixels in NTL images with a constant blooming distance 3.5 km. In fact, the blooming distance is related to the land covers and the brightness of artificial lights. This assumption leads to large errors in cities that have blooming distance different from 3.5 km. To address this problem, this study proposed an improved SEAM model (iSEAM) by considering spatial heterogeneity of blooming distance. Specifically, iSEAM segmented the DMSP-OLS image to obtain light objects and then employed the random forest method to estimate the effective blooming distance for each light object, and then corrected the blooming effect of all pixels in each light object by a modified pixel brightness interactive model. The test in China shows that the blooming distance ranges from 0 to 12.55 km in China, with an average 3.36 km. The correlation coefficient (R) between the images corrected by iSEAM and the NPP-VIIRS images reaches 0.70 that is higher than other blooming effect correction methods. Moreover, the corrected images by iSEAM have higher spatial heterogeneity than other methods. These results suggest that by considering the spatial heterogeneity of effective blooming distance, iSEAM can serve as a more accurate and effective method to correct the blooming effect of DMSP-OLS NTL images.