论文部分内容阅读
针对现有遥感影像变化检测方法常存在的检测结果破碎、虚检较多、对数据匹配要求高等问题。提出了一种融合像素级和对象级的遥感图像变化检测方法。利用光谱和纹理信息构建单高斯模型,在多尺度上进行像素级变化检测。然后,以像素级检测结果为种子区域,同时在变化前后影像上区域生长,融合生长结果提取变化对象。最后,依据检测需求对变化对象进行特征分类并滤除虚警。实验结果表明,该方法降低了虚检,保持了变化区域的结构完整性,在变化前后图像分辨率存在一定差别时仍有较高的检测精度。
In view of the existing remote sensing image change detection methods often exist in the detection of broken, more virtual inspection, data matching requirements higher issues. A new method of detecting the change of remote sensing images at pixel level and object level is proposed. Single-Gaussian model is constructed using spectral and texture information to detect pixel-level changes at multiple scales. Then, taking the pixel-level detection result as the seed region, the region of the image is grown before and after the change, and the change object is extracted by the fusion growth result. Finally, according to the detection needs of the characteristics of changes in the object classification and filtering out false alarms. The experimental results show that this method reduces the false detection and maintains the structural integrity of the changing area. When the image resolution is different, the method still has higher detection accuracy.