The Studies On Land Cover Change Detection Based On Object-oriented Method

Land cover change detection is the process that determines land cover change which is based on the same object or phenomenon observed in different time. It is a theory and method that extract and describe the time-varying characteristics of interested objects or phenomena. It provides quantitative analysis and determines the change which is based on the remote sensing image and the other reference GIS data in different time. GIS and RS as the main tool for it to extract dynamic information from multi-temporal data of the Earth’s environment and human activities, and its results are often used in the quick update of the GIS spatial database and spatial data mining.With the rapid development of geographic information system and remote sensing technology which trend of "three-multi"(multi-sensor, multi-platform and multi-angle) and the "three-high"(high spatial resolution, high spectral resolution and high temporal resolution), land cover change detection has been an active research field in remote sensing. Object-oriented image analysis technique (OOA) which is based on operating unit is a new geosciences analysis tools with a combination of RS and GIS, it has been extensively and successfully applied to the remote sensing image analysis such as image classification. It has been a favor of more and more searchers who major in land cover applications.Based on the analysis and summarizations of researches home and aboard, the concept of change detection was given, conceptual models and steps were summarized. In the dissertation, existing detection methods and basic principles of commonly used methods were expounded; their virtues and shortcomings were analyzed respectively. Then, this paper studied the related concepts and theory of object-oriented image analysis. The author focused on the multi-scale image segmentation technology and discussed how to determine these two important parameters, which are the involved bands of multispectral image and the weight of each involved band of multi-scale image segmentation in the eCognition of object-oriented remote sensing image analysis software. After that this paper made use of the object-oriented remote sensing image analysis technique in the land cover change detection with TM images, and compared the performance indexes between pixel-oriented change detection and object-oriented change detection, then the author analyzed the pros and cons of the three commonly used change detection methods and advantages of object-oriented change detection. Then we combined with the development trend of change detection, we advanced a combination with LBV transform and change vector analysis (CVA), which is called LBV-CVA analysis. The experimental results of the LBV-CVA analysis showed that the method effectively improved the detection accuracy and reduced the detection error. Finally, this paper summarized the main research results, pointed out the existing problems and forecasted the future research works.