Research Of High Resolution Remote Sensing Image Segmentation Based On Modified Watershed

Nowadays,with the continuous improvement of spatial resolution,the information on the high resolution remote sensing image has been enriched a lot. People can get more and more detailed observational data of the earth through the high resolution remote sensing satellite. The development has greatly inspired the appearance of various applications in many fields for remote sensing technology. However, on the other hand new challenge has been brought to the image processing technology because of the large amount of information and complicated noise from background.As the foundation of remote sensing information extraction and automatic recognition,Image segmentation technology has became the most urgent problem to be improved.In view of the problems such as large amount data, complex texture details background noise, different spectra characteristics with the same object and the opposite condition, a study of high resolution remote sensing image segmentation based on modified watershed transformation is conducted, making it suitable for high resolution remote sensing image.The paper introduces a kind of mixed opening and closing reconstruction filter based on mathematical morphology, which can smooth the texture details and get rid of complex background noise. On the other hand, with the fully consideration of geometry and size features, the filter provides high value of reference for image smoothing theory. In addition of extreme value marking algorithm and the reconstruction of gradient image, a remote sensing image segmentation algorithm combined with morphological filter and marked watershed transformation is made up.The algorithm can restrain the over-segmentation phenomenon of traditional watershed transformation and improve the precision of image segmentation.According to the characters of multispectral remote sensing image, this paper introduces a kind of gradient calculation method on RGB vector space. Combined with self-adaptive extended minimal transform technology, the accuracy of image segmentation results is improved further. For the evaluation of segmentation results,this paper introduced image segmentation accuracy evaluation method of inconsistency. An analysis between segmentation results and the ideal results is superimposed on calculating the algebraic and geometric differences. Therefore six evaluation indexes are got by the analysis. In addition to the evaluation, At the last,the evaluation of accuracy of this study was compared with the results in e Cognition.