多特征融合的遥感影像河流提取方法

针对遥感影像河流识别率较低问题,文中提出一种多特征融合的遥感影像河流提取方法.该方法首先利用多尺度分割算法将影像分割为不同的对象,提取影像对象的水体指数、阴影水体指数、局部纹理与颜色等特征,利用极限学习机进行训练和识别;然后,对极限学习机的粗检测结果利用多判据软投票法优化获得最优的水体检测结果;最后,利用高分二号数据进...

River extraction method of remote sensing image based on multi feature fusion

In view of the low recognition rate of river in remote sensing image,in this paper,the river extraction method was proposed based on the multi feature fusion.Firstly,the multi-scale segmentation algorithm was used to segment the image into the different objects,and the characteristics of the water index,the shadow water index,the local texture and the color of the image object were extracted,and then the extreme learning machine was used for training and identifying;then,the optimal water detection results were optimized and obtained using the multi criteria soft voting method for the rough detection results of the extreme learning machine;finally,the Gaofen-2 data was used to experimental,and the detection accuracy was more than 94%,and the experimental results showed that the river could be effectively extracted by the method in the complex background.