Present Situation and Trend of Remote Sensing Land Use/Cover Classification Extraction

To further advance the automatic process of land use/cover (LULC) classification extraction through remote sensing (RS) images, by reading many literatures, we summarized the problems, research difficulties and development trends in the process of information extraction and classification of land use. Overall, LULC Classification and extraction based on RS images include 3 tasks: data source selection, sampling design, classification method selection and classifier performance evaluation. These tasks are all important, that is, interdependence and mutual influence. The OBIC method has become a popular method of L ULC classification because it makes full use of geographic information system (GIS) technology to process spatial, spectral and textural features in RS images. There are many OBIC algorithms, especially the Machine learning (ML) algorithms offers the potential for effectiveness and efficiency, such as Random forest (RF), Support vector machine (SVM) and so on. The Object-based image classification (OBIC) method involves three stages: segmentation, feature-selection and classification. A large number of studies have proved that there are many problems in each task of the LCLU classification extraction method based on RS images. These problems include design of sample sampling strategy, determination of optimal image segmentation parameters and optimization of parameter of classification algorithm and so on. At present, solving these problems requires frequent human-computer interaction also has a great negative influence on the automatic extraction process of remote sensing classification. U sing GIS technology to promote the automatic extraction of remote sensing classification has become a trend of the development of remote sensing classification method.