Research And Implement Of Information Extraction On The Road Under High-resolution Remote Sensing Images

With the application and research of high-resolution remote sensing images, there are more and more information in remote sensing images. High-resolution images of remote sensing has played important role on land survey, mapping, urban planning,aviation and national defence. However, in terms of pattern recognition of remote sensing images, the improvement of space resolution is not equal to the improvement of classification. This is because that with the improvement of resolution,there are more and more details and small objects distinct that will produce many same things and heterogeneity, in the internal structure, and will also reduce the difference of the spectrum. Therefore, it only depending on the image of the spectrum is obviously, and there will be categorized more of points and wrong.By the data from Wenchuan area,we used SVM to extract damaged roads from remote sensing images.At present,with the continuous perfection of wavelet,the paper I wrote would put forward a improved algorithm on wavelet to get rid of interferential information and strengthen correlative information in remote sensing image. By large numbers of experiments, the result that would provide better platform for next process has more obvious improvement than other traditional ways. SVM as a new study based on statistical theory of classification algorithms, which is compared with other classification algorithms, it has many advantages,such as the global optimal, simple structure and promotion of advantages of strong ability. In the end,the paper will construct a model of adamaged road to estimate the extent of adamaged road.The experiment results show that SVM &the improved wavelet method has a good better performance than other ways,such as parallel lines, the threshold, k-means, and object-oriented.