Potential Analysis Of CBERS-02B CCD Imagery Used In Large Area Land Cover Mapping

The CBERS-02B is a real-time transmission Earth observation remote sensing satellite. It is the third of the resource satellite series developed by China in cooperation with Brazil. The CBERS satellites provide continuous and stable Earth observation data for China and Brazil, which has been widely used in many fields and played a significant role in national economic construction and social development. The major research of this paper is to analyze the potential of CBERS-02B CCD images used in large-scale land cover mapping. The main aspects of potential analysis include the evaluation of the capability of large-scale mapping, the geometric and radiometric quality and the classification capability. This thesis is organized as the follows: (1) introduction of the basic features of CCD images, such as spatial resolution and width, band settings, and temporal resolution. These features are of importance to large scale land cover classification. (2) analysis of the geometric quality of systematically corrected CCD image, such as band to band registration accuracy, absolute positioning accuracy and registration accuracy of multi-temporal CBERS-02B images. The analysis result showed that the geometric quality of CBERS-02B image was not very high, which imposed difficulities on image mosaic processing to cover large area. (3) analysis of the radiometric quality. Several parameters, such as entropy, signal to noise ratio and definition were calculated to evaluate the radiometric quality of CBERS-02B image. We found that the fourth band had relatively better radiometric quality than the first three bands. (4) evaluation of classification capability. The classification performance of CBERS-02B was compared with that of Landsat ETM+ image. The ISODATA unsupervised classifier and maximum likelihood supervised classifier were used to. classify the CCD and ETM+ original bands and images joined with NDVI index respectively. The classification results showed that the ISODATA classifier used in CCD and ETM+ images got a lower accuracy. But the classification result of CCD image joining with NDVI index used maximum likelihood classifier got a higher total classification accuracy. And the author applied the decision tree classifier in CCD land cover classification and got a total classification accuracy of 87.42%. (5)Finally, the author tried to build a model and criteria of the potential analysis, and used the fuzzy comprehensive evaluation method to analysis all the indexes and the result showed that: 02B CCD imagery has potential as the data pool of large-scale land-cover classification.