Study On Lithological Identification Of Remote Sensing Technology Based On Object - Oriented

Due to development of last several decades, remote sensing technology are more and more mature in geological field. The computer automatic classification technology, especially the image classification method based on pixels has become an important mean for lithology identification. However, more research show that there’re still some problems about classification methods based on pikels. With the improvement of remote sensing data resolution, the surface features of image such as texture, structure and spatial relationship has involved in lithology identification study gradually. On this background, the object-oriented image analysis technology has been given considerable attention and gradually been in a hot research of the remote sensing lithology identification. This paper relies on the high-tech industry development project plans of Yunnan province "The application of mineral resources remote forecast evaluation technology in Yunnan Northeast mine set district" and school-enterprises cooperation project "Huize super-large lead-zinc deposit of deep and periphery of concealed ore prospecting prediction and increase reserve research", taking Huize lead-zinc mining area of Yunnan as an example, and has made a contrastive study of lithology classification methods based on pixels and object-oriented by using the high resolution remote sensing imagery, which provides technical reference for rapid and accurate lithology identification by using the remote sensing technology.On the basis of fully understanding the rocks, tectonic and other geological background and typical rock and mineral spectrum characteristics in the study area, this research through a lithology identification comparative analysis of the carbonate rocks, basalt and sandstone by using RapidEye multispectral image, and means of supervised classification and object-oriented classification, shows that the object-oriented classification method is more conducive to lithology classification and identification and has high classification accuracy, by making full use of the rock spectral information and spatial information such as texture, geometric feature and location. In the end, analyzing landscape pattern of the object-oriented lithology classification graph, the landscape features of basalt and carbonate rock area are contrastive analyzed in a quantitative manner. The study has a certain guiding significance for lithology identification fast and exactly by object-oriented high-resolution image classification technology and in-depth understanding the corresponding relation between the lithologic unit landscape characteristics and their remote sensing visual interpreting marks.The main results of this paper as follows:(1)The multi-scale segmentation of image is conducted by object-oriented image analysis technology which comprehensively utilizing the spectral and spatial information of the high resolution image and in the principle of minimizing the spectral and spatial heterogeneity of ground objects. As a recognition unit, the object is better to express the ground objects’attributes than pixels, and the object-oriented classification results are more in line with the needs of practical application.(2)Using methods of supervised classification and object-oriented classification to classify the ground objects in study area, and the results indicated that the accuracy and classifying quality of object-oriented image analysis technology is the highest, and the completeness of ground objects is better too. Among the supervised classification methods, the accuracy and classifying quality of maximum likelihood, neural network and support vector machine is comparatively better than others, but there are some leaking classified points and sub-phenomenon mistakes.(3)Comparative study of lithology identification found that there are many’salt and pepper’phenomenon in the lithology recognition results by supervised classification, and it’s hard to circle the complete lithologic units with poor continuity. While the oriented-object lithology classification results have the highest accuracy, with more clear lithologic boundary, accurate position, better integrity and spatial continuity, and high degree of agreement with the geological lithologic unit. The latter also effectively restrained the interference of vegetation and shadow to basalt.(4)The results of landscape pattern analysis in lithologic classification map show that the fragmentation degree of basalt rocks is higher than the carbonate rocks and its geometric features appear irregular clumps, while the carbonate rocks show a strip or sheet distribution characteristics. The analysis results are in agreement with the degree of weathering, outcropped situation and the graphic character of remote sensing image of carbonate rock and basalt area.