The Scale Effect Analysis In Identification Of Forest Types From Multi-resolution Remote Sensing

Since Remote Sensing technology obtained very good development and wide application, the main data source used to study geographic phenomenon has gradually transferred to Remote Sensing images. Spatial resolution is one of the most important attributes of Remote Sensing images, namely measurement scale. On one hand, the basic characteristics of remote sensing data is determined by the changes of geographical phenomena nature scale; On the other hand, the changes of spatial resolution can reflect the properties of land surface more accurately at different scales. The scale dependence basically exist in the geographic phenomenon, so the properties observed, as well as some principles or laws resulted from this in a single scale may be effective, similar or even need to be modified in other scales.In the process of remote sensing data information extraction, the basic characteristics of scale make its role be payed more and more attention. Meanwhile, in remote sensing, for the same object studied, different staffs use the dates of different scales or different spatial resolution, which leads to conclusions disparities.Changing scale rises to restrict action on information observing, indicating, analyzing and exchanging, so we have to understand how its changing characteristics change as the change of scale, when we describe a certain phenomenon and the process of change in a single scale. The scale of the problem of remote sensing information extraction in remote sensing applications How to select the appropriate scale image, Reduce the extraction of information uncertainty, Reduce the blindness of data selection played a significant role as well as comprehensive utilization of multi-scale information.This thesis is10meters SPOT5,19.5meters of CBERS,30meters TM多光谱multi-spectral data for the study, Outside the industry measured point data and2008forest map data as reference source data to research in remote sensing scale problems, specific contents and conclusions include:(1)Different resolution remote sensing data preprocessed. Data preprocessing included band selection, geometric correction, image enhancement and processing the crop. At the band selection, SPOTS the band combination for band4. band3. band1: then SPOTS. CBERS. TM多光谱multi-spectral data conduct geometric correction and corrected accuracy results were as follows:0.5.0.7.1.0; Image enhancement used edge enhancement. (2)SPOT5data scale expansion. Scale converted to the data or information from one scale to another scale. SPOT5data for the study, using the simple average method, nearest neighbor sampling and maximize retention the France to expand the image. Extension method and select a suitable method to assess the scale of the mean, standard deviation and area differences. The analysis shows that: the simple average method is superior to the nearest neighbor sampling. Using the simple average method as SPOT5multispectral images scale expansion method.(3)The scale of the different forest types to select. Geographical phenomena of a particular type of research, you should use the appropriate image data. In the analysis the multi-scale surface phenomena, multi-scale remote sensing data is the main source the data. However, for some research purposes the relatively clear, choose what kind of scale remote sensing data is more conducive to this type of recognition is very important. The results show that the higher the resolution, the higher the recognition accuracy of forest type is not necessarily; The same resolution of different data sources, the recognition accuracy of each forest type is not the same; Identification of the different forest types, the optimal resolution; King for a multiple forest types, a unique optimal resolution is suitable for the identification of all forest types, but there is a suitable space for a particular forest type resolution range.Analysis of multi-resolution remote sensing images of forest types recognized by the scale effect can not only improve the accuracy of the identification of forest types, and also provides a reference for the investigation of forest resources and practical application.