Classification Algorithm For High-resolution Image Of Mining Based On Multi-Scale Rule Sets

China is rich in mineral resources and the mining area planning and supervision is not only very important duty for mineral resources departments,but also great significance for the development and management of mineral resources of our country.Platforms and sensors are used in remote sensing technology relying on the continuous development and improvement of spatial resolution,and remote sensing data acquired is improving.These HR data accurately reflect the mining feature of shape,size,and characteristics of texture and topology relations.How to quickly,accurately and automatically make the extraction and use of the available information in the data has become an urgent problem in the field of remote sensing information extraction.The phenomenon of different spectra characteristics with the same object and different objects same image is serve in mining area,and other shape features as texture and structure is obvious.Traditional method of information extraction based on pixels,meet now to obtain high resolution remote sensing image,it is difficult to make full use of the high score of remote sensing image has the texture,structure and other information,which can lead to serious phenomenon of "salt and pepper".In the high score of remote sensing image information extraction and classification using object-oriented classification method can get good effect,and the method because of its high precision and strong pertinence,advantages and widely research and application.The most commonly used segmentation process of object-oriented classification method is a multi-scale segmentation,but this segmentation method is easy to result in"over-segmentation" and "less-segmentation" phenomenon because of measure problems.Classification method generally use the classification based on rule set,but the establishment of the rule sets need to keep trying different characteristics and corresponding threshold to find the most appropriate rules,it often requires a lot of manpower and time costs.Aiming at the above problems,1)proposed a segmentation method based on multiscale edge constraint,will feature edge feature EDISON obtained as a multiscale segmentation process constraints parameter,reducing the "over " and "less-segmentation " error;2)proposed S-Ant-Miner rule terrain classification method,SEaTH algorithm obtained features and the corresponding threshold as the Ant-Miner algorithm decision attributes and breakpoints.Then follow the next steps Ant-Miner classification rule mining algorithm,which optimized the classification rules set,enhanced the rule mining efficiency,improved the classification accuracy.Finally,this paper applied the improved object-oriented classification method in GuanBaoShan and DongAnShan iron ore high marks in the feature classification.The stope,tailings DAMS,roads,water,building,bare land,vegetation in the image such as feature extraction and classification categories and produced the mining classification thematic map.On the basis of the classification results of GuanBaoShan image has carried on the vegetation change detection and roads.The object-oriented classification method are put forward in this paper could improve classification accuracy and efficiency and provide more accurate basis for mine management department materials,and also has a certain practical application value.