Study On The Extraction Of Major Geological Objects In The Best Segmentation Scale

With the rapid development of science and technology,the development of remote sensing platform and sensor technology is changing rapidly.Object oriented analysis as a bridge connecting the remote sensing and geographic information system,which occupies the dominant position in the high score image extraction classification application,and has a traditional pixel based classification method can not match the huge advantage,is also the primary means of geographic information extraction in high resolution image.However,in fact,objective things have the characteristics of multi-level and multi-scale.The phenomena and objects in nature can only be observed in a certain scale.The same object at different scales will also show a completely different landscape features.Using the same segmentation scale to separate the results of different objects will also be different.Therefore,it is important to distinguish the different objects with different segmentation scales in the high resolution object-oriented analysis.This paper focuses on the discussion for different objects,to select the optimal segmentation scale to distinguish the object itself based on the inherent characteristics of the object,and the final result with the traditional pixel based classification method of comparative analysis,and gives the accuracy evaluation.The main results of this paper are as follows:(1)On the basis of a large number of literature on related research,according to the characteristics of image data to be processed,for acquiring the image object classification method,the mean variance method,the maximum area method and vector distance method to determine the optimal segmentation scale.Based on the graph of the corresponding scale and the complexity of the actual objects,the optimal segmentation scale is selected to create different scale object layers.Experiments show that the optimal segmentation scale selected by the above method can distinguish different objects well,effectively avoiding the "over-segmentation" and "under-segmentation",so that the image object separability greatly improved.To a large extent improve the accuracy of high score image classification extraction.(2)Based on the established multi-scale layer,aiming at the difficulty of information extraction and different characteristics of the main city of the object,the object hierarchy network method has been designed.Firstly,the classification rules are set up according to the spatial and geographical characteristics,the characteristics of the spectral features,the geometric features,the texture features and the class-related features.Based on the classification method,such as threshold method,membership function method and nearest neighbor classification method,the classification of urban features in remote sensing images of Dujiangyan urban area is classified to extract the roads,houses,waters,woodland,bare land And other major features.As the vegetation and building shadows will cause great interference to the extraction of the road,in order to improve the extraction accuracy of the road experiments proposed a shadow compensation method,regional growth method and other methods,effectively reducing the road extraction error to improve the road classification extract accuracy.To provide the basis for the rapid extraction of urban features in high score remote sensing images.(3)Compare the traditional image-based classification method and object-oriented multi-scale segmentation method to classify the same image and final classification results to prove that Multi-scale object-oriented segmentation method has a significant advantage in extracting high-value image city features.