Based On The Artificial Surface Features Of Object-oriented High-resolution Remote Sensing Image Information Extraction

In recent years, with remote sensing technology rapid development and the construction of"digital city"the quickening of the process, how to extract high-resolution remote sensing image of Man-made objects in remote sensing research has become the key and difficult point. At present, many experts and scholars at home and abroad for urban Man-made objects extraction has done a lot of research, has made a lot of achievements and experience. However, High resolution of remote sensing image extracted urban Man-made objects is a relatively frontiers and new field of research, currently used in production of the method is not mature, and basic in experiment and theoretical research stage.The traditional remote sensing image of Man-made objects extraction method is artificial based on pixels, This method has many shortcomings such as classification accuracy, low video information not make full use of, extracting slow, etc. This paper put forward based on the object-oriented method is used in high resolution man-made object in the remote sensing image the extraction of high resolution remote sensing images, according to the characteristics of the abundant information, This paper presents a method based on sample standard deviation model to determine the best segmentation scale used in many class features and extraction. This paper selects the Liaoning Donggang, Guangxi Guilin and Guizhou Zhunyi remote sensing data of high resolution, using eCongnition based on object oriented high-resolution remote sensing image information extraction using artificial ryzhkov trial included buildings, roads, Bridges. The extraction of experiments in building is proposed based on Roberts edge detection method to estimate the segmentation scale, the speed of the choice of parameters is improved. Finally using classification accuracy, Kappa coefficient for the extraction of the results, the results showed that the quantitative evaluation based on object-oriented method suitable for high resolution in remote sensing image extracted, and have artificial features high efficiency and precision.