Research On Vegetation Information Extraction Using High-resolution Remotely Sensed Image Object-oriented Method Based

In recent years to understand and grasp the trend of vegetation cover change hasbecome a growing focus of attention, many researchers have done a lot of problemsassociated with vegetation coverage, vegetation resources to reasonable, effective andefficient management, so as to promote coordinated development of resources andenvironment and social economy. This thesis mainly focuses on the study ofvegetation information extraction, the study of vegetation classification is weak atpresent, classification accuracy is not high, this article selects a north Korea as theresearch area, using the object-oriented method to study the vegetation classificationmethod, thus improve the vegetation classification precision.In this paper, using high resolution remote sensing image of GeoEye-1, basedon the object-oriented image classification method, to extract information ofvegetation in the study area, the main results are as follows:(1)In the study of first according to the geological characteristics and adopted bythe data source, data pretreatment such as geometric correction and image fusion.Also in the study on image fusion technology has carried on the thorough analysis,put forward the combination of wavelet and IHS way, improve the quality andprecision of the fusion image, lays the foundation for the later vegetation informationextraction.(2)By multi-scale segmentation experiments, this paper discusses the problem ofthe scale of the image segmentation and its parameter selection problem, got theoptimal scale based on the different characteristics of the image. In determining thedifferent features on the basis of the optimal scale, the use of object-orientedclassification method, respectively, according to different features of the optimal scale,high resolution remote sensing image is roughly divided into woodland, grassland,building, bare land, the river5classes, and the results of its use of confusing momentis precision evaluation, precision as high as90.14%, kappa coefficient of0.8514;(3)At the same time using the remote sensing image classification method basedon pixels, supervised classification was carried out on the same area of the image,unsupervised classification, and the decision tree classification, accuracy evaluation;Results show that the object-oriented classification method of classification accuracyis higher than the classification method based on pixels, as a result, the research ofusing the adjacent classification and fuzzy classification method respectively to forestland and grassland in the subdivision, realized the high resolution remotesensing image of vegetation information extraction.