IMPROVEMENT OF CLASSIFICATION ACCURACY IN AIRPLANE BOUNDARY BY USING SMOOTHING TECHNIQUES

摘要

We have developed some algorithms for airplane detection from high resolution satellite image. High resolution satellite image has two kind of images. One is panchromatic image. The other is multi spectral image. Multi spectral image has color information. But, ground resolution of multi spectral image is less than that of panchromatic image. Pan-sharpen image was created to improve the ground resolution of multi spectral image. We tried to classify the pan-sharpen image for detection of airplane. Classified result showed some jitters in boundary area of airplane. These jitters came from low resolution of multi spectral image. In this study, we proposed an algorithm to improve the classification accuracy in boundary area of airplane by using smoothing techniques. Airplane detetion was conducted by combining the classified result and some geometric features. We showed the effectiveness of the proposed algorithm.