Scenic Resource Meticulous Extraction Based On Multi-Source Information Fusion

The scenic area is an important part of the natural heritage of the country and contains a lot of historical and cultural information. In order to avoid over-exploitation of the natural,degradation of cultural landscape, it is necessary to dynamically monitor the illegal construction and land abuse of the scenic area. Remote sensing images are widely used in various fields of social life and national economy because of its inherently unique advantages.Especially the appearance of high spatial resolution remote sensing images made close observation of the Earth's surface available. In this paper, high spatial remote sensing image and DEM(digital elevation model) data are combined to extract the scenic resource for scenic monitoring, evaluation, planning and management objectively and accurately.In view of the particularity of scenic resources(including surface coverings and mountain), different targets are extracted respectively with 1m remote sensing image and DEM because of their own characteristics. Target recognition process through remote sensing image is: First, take geometric precision correction and radiometric calibration steps for image pre-processing. Secondly, the paper proposes a parallel computing multi-scale segmentation approach that can not only improve the recognition accuracy, but also greatly reduced recognition time. Thirdly, based on the widely used SVM, the maximum likelihood and K-means classification analysis, the paper proposes a simple and effective supervised classification technique that is based on the model of multi-feature recognition algorithm, and the relevant comparative experiments verification. On the basis of comprehensive extraction,analysis, experiments, color difference, texture, histogram and variance functions as the characteristics of the recognition factor to model. Finally, use the shortest distance method as a recognition criterion to obtain a final recognition result, and the recognition resolution researches up to 85%. The DEM target identification process is: combine linear analysis window and Split method to extract terrain feature line. The experimental results show that the proposed method has a good application value.