Landsat Tm Imagery Classification Using Object-oriented Technology

For recent20years, various types of satellite remote sensing have widely been used in the land utilization, environmental monitoring, covering agricultural output assessment, forest fires and government decision-making, and so on. On most of the situation, the use of remote sensing data need classify for classified data (project data), classification technology is also called the image analysis technique.The classification of remote sensing image data is always based on the pixels. Project information which based on the pixels of remote sensing image classification method extracted have much error, not can reflect all the information of the remote sensing data contains, such as spatial relations information, etc.So produce many of the invalid broken figure spot, which reduces the accuracy of classification. To solve these problems, a recently developed new image classification method-the object-oriented classification, which is based on image Object classification method.Image object is point to the collection of pixels that have certain similar characteristics. Compared with the traditional remote sensing image classification method, object-oriented method has many advantages:Image segmentation can separate image object constituted by the same properties of pixels and the differences of spectra value within the objects are extremely small and even can be ignored, for this reason, the "salt and pepper phenomenon" can be eliminateed;Using shape, space and texture feature information of remote sensing image objects to classify reduced the influence caused by "the same object but different spectrum" and "the same spectrum but different object" of classification accuracy; Adopting multi-scale segmentation method can meet the requirements of extracting features of different ground object.The main purpose of the paper is to classify the TM images of Erhai lake basin at Dali Yunnan province using the object-oriented image classification method. Extracting information of Erhai lake terrain can provide the basis for using the resources of region reasonably.The experiment used the object-oriented image classification method whose Experimental study platform is the commercial software eCognition Developer8.0and object of study is the Landsat TM satellite images of Erhai lake basin. Feature information of Water body, residents, forest land, dry land, Luodi, farmland, road grassland and the shadow of Landsat TM satellite images have been extracted successfully by using this method and high precision classification results have been obtained, which has important and profound significance for city planning, environmental protection and ecological protection and pollution control.This article systematically analyzed the method of object-oriented image. This method not only can extract Landsat TM of moderate resolution satellite image features information, but also can be used to extract high resolution QuickBird (QB). IKONOS, World View, GeoEye satellite remote sensing image features information and gain high accuracy of classification. This paper did some practical significance of technology work, which have an important reference meanings on the classification of moderate resolution of remote sensing satellite images.