基于超像素分割变化特征提取的高铁环境变化监测

High Speed Railway Environmental Change Monitoring Based on Super Pixel Segmentation Change Feature Extraction

为了保障高铁安全有序运营,需要定期进行高铁沿线运营环境的变化监测,及时发现对高铁具有威胁的地表目标。采用“高分二号”遥感数据作为数据源,使用改进型多尺度Forstner算法进行特征点提取,完成不同时相影像在空间上的精确匹配。使用SLIC超像素分割算法进行影像对象分割,将影像数据由单一像素转化为地物对象。对地物样本对象单元进行统计,确定分类阈值,进行变化区、未变化区、混淆区域的分类,再综合利用两个时相的NDVI、NDWI和亮度特征信息对混淆分类区域作进一步分类。在变化区、未变化区各随机生成100个检验点,通过对检验点的统计可知:该方法总体精度为91%,有效抑制了“椒盐噪声”,使得变化区域边界与真实地物状况更加吻合。

In order to ensure the safety and orderly operation of high-speed rail,it is necessary to regularly monitor changes in the operating environment along the high-speed rail to timely discover surface targets that are threatening to high-speed rail. The high-resolution No. 2 remote sensing data is used as the data source, and the improved multi-scale Forstner algorithm is used to extract the feature points to complete the spatial matching of different time-phase images. The SLIC super pixel segmentation algorithm is used to segment the image object,and the image data is converted from a single pixel to a ground object. The sample units of the ground object are counted, the classification threshold is determined, the change area, the unchanged area and the confusion area are classified,and the confusing area is further classified by using the NDVI,NDWI and brightness characteristic information of the two phases. 100 test points are randomly generated in the change zone and the unchanged zone. According to the statistics of the test points,the overall accuracy of the method is 91%,which effectively suppresses the“salt and pepper noise”,making the boundary of the change region more consistent with the real ground feature state.