Based On Multi-feature Classification Of Urban Remote Sensing Dynamic Monitoring

With the continued development of the economy, the pace of urban construction is accelerating. Timely access to urban development information, including the development trend, is the foundation of proper planning, construction and management of urban development. Remote Sensing using dynamic monitoring approach has the characteristics of objective, fast and can cover an extensive area. These features offer unparalleled advantages in urban resources exploration over conventional means. This approach adopts digital image processing technology to conduct exploration and monitoring of sensor data of urban remote sensing images. After a series of data processing, including fusion, principal component transformation, algebraic operation and image mapping, of remote sensing images of different phases, various ground objects which have changed can be automatically distinguished from complicated surrounding environment. The processes and methods for dynamic monitoring remote sensing information can be varied due to the differences in data sources and the processing targets. However, during the processing of the data, the accuracy of the classification of information and the changed information can determine the reliability of the results of dynamic monitoring.This paper conducts research on the extraction method for urban remote sensing base information and the commonly used remote sensing dynamic monitoring methods. Based on the remote sensing images of urban area of Kunming acquired in 2003 and 2008, this paper proposes an urban remote sensing dynamic monitoring method which integrates multiple feature classification comparison methods to accomplish dynamic monitoring of remote sensing information on Kunming.To meet the remote sensing information extraction requirements, this paper undertakes research, as well as processing, on various issues, such as geometric registration, shadow correction and image enhancement during the pretreatment process of remote sensing images. Besides, this paper generates a model which is based on shadow processing method to implement shadow correction to images. Furthermore, the paper also conducts research on the extraction method of the four types of major base information, namely, vegetation, water body, trunk roads and buildings in remote sensing images of urban area of Kunming. This paper proposes an approach that combines erosion as well as opening and closing operation based on mathematical morphology with structural characteristics of the road to extract urban trunk road information, and uses exclusive method of classification information to extract areas of buildings, resulting in higher accuracy in information extraction and changed information extraction. Furthermore, this paper, adopts principal component analysis and multiple feature classification methods proposed in the paper to perform dynamic monitoring of remote sensing images of urban area of Kunming, compares the process effectiveness as well as analyzes the characteristics and practicability of the two approaches.The experiment results of the comparison indicate the process method proposed has achieved good results in periodic dynamic monitoring of urban remote sensing information. Finally, the paper identifies areas where improvements can be made.