Research On Land Use Change Detection Based On Data Mining Technology

The land use change survey is one of the important survey which needs to be carried out every year in China.The information of urban land use change is of great significance to the analysis of the city's economic development,the monitoring of urban land use change and the analysis of urban natural ecological environment.The aerial remote sensing image data has the characteristics of high spatial resolution,real-time data acquisition and low cost,which is a good data source for urban land-use change detection.At present,there are some problems in urban land use change detection by using aerial remote sensing image data,such as data preprocessing methods selection,low resolution of spectral resolution,the phenomenon of"homologous dissimilarity",and the low degree of automation of land use change detection and so on.To solve these problems,the C4.5 algorithm based on data mining was applied to the decision tree classification rule generation,and was used to the studies ofthe land use change detection method.Compared with the land use change detection results based on the object-oriented classification method,the following conclusions were obtained:(1)In the pretreatment process of aerial remote sensing image data,the automatic matching based on gray information and artificial selection method were adopted to the Image Registration progression,and the sub-pixel level registration accuracy could be achieved;Radiation correction of aerial remote sensing image by means of mean-standard deviation relative radiation normalization method could achieve high similarity in front and back time image,and the least amount of information loss.(2)The classification errors caused by the phenomenon of "homologous dissimilarity" could be effectively weakened by controlling the quality of sample collection and training in high-resolution aerial remote sensing image classification.At the time of collecting samples,the classification of some objects with the phenomenon of "homologous dissimilarity" was refined,and the classification precision could be improved effectively.The training quality of samples could be evaluated by J-M separability index.(3)The feature space including image spectral information and texture information was constructed,the effectiveness of texture information in land use change detection was validated,and maked up for the lack of spectralr solution of aerial remote sensing image data,and maked the data more fully used.(4)The decision tree classification method based on data mining C4.5 algorithm was applied to land use change detection,and the classification accuracy and change detection precision were equivalent to those based on object-oriented method.The results showed that the land use classification of decision tree based on data mining C4.5 algorithm was superior to object-oriented method in artificial dependency,and it could also improve the work efficiency and guarantee the accuracy of classification.