基于小波核双重加权SVM模型的蓝藻水华识别与变化检测

Recognition and change detection of cyanobacteria bloom using the improved SVM model based on the wavelet kernel and dual weight

以淀山湖为研究区域,利用Landsat系列遥感影像,提出了归一化蓝藻指数的构建方法,结合波段组合和Gabor滤波器构建多特征空间,并基于变精度粗糙集和灰色关联决策相结合的方法进行特征空间的优化。在此基础上,采用小波核双重加权SVM分类模型,得到研究区蓝藻水华空间分布格局的识别与检测结果。以误差矩阵证明此分类模型能够较准确地识别出蓝藻水华覆盖区,满足环境调查的要求。研究成果为淀山湖蓝藻水华的防治和水生生态系统的保护提供了科学依据。

Taking Dianshanhu Lake as the study area, the normalized difference cyanobacteria bloom index （NDI_CB） was developed to identify cyanobacteria blooms in remotely sensed Landsat images. Multi-feature space obtained via band combination and a Gabor filter was optimized based on the variable precision rough set （VPRS） and grey incidence decision （GID）. Using these data, the improved support vector machine （SVM） classification model based on the wavelet kernel and dual-weight was applied to extract spatial distribution information and change results for cyanobacteria bloom in the study area. The results show that the accuracy of the classification model could be improved, and it could also meet the requirements of environmental investigation. Furthermore, it demonstrated that the results could provide a scientific basis for the prevention of cyanobacteda bloom and protection of the aquatic ecosystem in Dianshanhu Lake.