Object-oriented High-resolution Remote Sensing Image Classification Of Land Cover

With economic development more and more people attach importance to land use, land for non-renewable resources, should be the people seriously. Traditional land cover survey, are time-consuming, costly, resulting in a tremendous waste of human and material resources, and remote sensing technology, with its access to information synchronization, timeliness, comprehensiveness and economic characteristics of the land cover information fast acquisition and evaluation of a new technique. In high-resolution remote sensing images for land cover classification study, detailing the key object-oriented information extraction-multiscale image segmentation and classification technique based segmentation, and spectral information with the traditional pixel-based classification methods were comparison. Based on the above theory, the experimental areas feature information extraction experiment, to achieve land land cover classification. Final classification results showed that:in the visual aspect of object-oriented information extraction technology to overcome the traditional methods can not be overcome, noise; on accuracy in terms of object-oriented information extraction technology overall accuracy than the maximum likelihood method to improve the overall accuracy, and the feature information extraction accuracy of various types are improved, which fields, roads, buildings accuracy.Main research contents:(1) According to the characteristics of different types of feature, select the feature to extract the optimal segmentation scale experimental zone segmentation, feature type extraction constructed classification system.(2) Types of feature extraction feature or combination of features, using the appropriate classification method to classify the type of feature, obtain the experimental area terrain classification result.(3) The experimental area of object-oriented classification method of remote sensing image classification results with the traditional classification methods for pixel classification results for comparative evaluation.(4) Using object-oriented high-resolution remote sensing image classification method for remote sensing image information extraction, higher accuracy.