The Object-oriented High-resolution Remote Sensing Image Classification And Debris Flow Information Extraction Research

Debris flow is one of the natural disasters widely distributed around the world and causes major loss of life and property every year. Therefore, the extraction of the debris flow from remote sensing image has a very important significance for the disaster relief and disaster reconstruction. Since the early visual interpretation in the high-resolution remote sensing image information extraction is widely used but time-consuming. The subject of how to use computer technology to extract the debris flow fast and accurately has a practical application.Compared with the medium and low resolution image, the high-resolution image has the richer and more obvious texture and spatial information. It is difficult using only the spectral information to achieve satisfactory results. And pixel-based classification methods appear to be prone to a "salt and pepper phenomenon". In order to solve the problem, based on the multi-features, the method of object-oriented image analysis can be used.The main contributions of this paper include:(1) This paper focus on the preprocessing of the remote sensing image and feature extraction. The preprocessing mainly contains the radiometric calibration, geometric correction, image fusion and image enhancement. Then, the feature selection and extraction of remote sensing image after pretreatment is carried out. These features include:the ratio index feature, normalize difference soil index feature, normalize difference vegetation index feature, texture feature of the different window size, PC A feature and ICA features.(2) This paper studies the extraction of the debris flow on pixel-based classification of remote sensing images. Based on spectral feature, texture feature, index feature, pixel-based classification method is studied. Then, based on full analysis of the texture feature with different image window size, the most appropriate window size of land cover classification and debris flow areas extraction was determined. The experiments show that the debris flow can be extracted effectively and accurately by using the spectral features, texture feature and index feature.(3) Based on object-oriented technology, we propose an object-oriented classification and extraction algorithm of debris flow. This method firstly select the best segmentation scale to split out the different sub-areas, then use the "voting rules" strategy to determine the category of sub-regional, and finally complete object-oriented remote sensing image classification. The experiments show that object-oriented classification method outperforms the pixel-based classification method.