Fuzzy Color Similarity And Its Application In Remote Sensing Image Processing

The high-resolution remote sensing image represented by the on-satellite multi-spectral remote sensing image not only reflects the unique spectral information of the object,but also accurately expresses the spatial information such as the position,shape and structure of the object.It is widely used because of its low cost,large coverage,high precision,and large amount of information.The rapid improvement of the quality of remote sensing data has created unprecedented advantages for practical applications.However,due to the difficulty of data correlation,fine scale,high dimension and complex structure,it has become the bottleneck of remote sensing image application.In order to solve the problems of strong correlation,complex structure and difficult sequencing,the thesis introduces the similarity theory into the color space,combines the fuzzy theory,the fuzzy similarity measure consistent with human visual perception is defined to measure the correlation between vectors,and construct the fuzzy color similarity measure framework based on the above.According to these,the color morphological operation suitable for remote sensing images is constructed,and the application processing of remote sensing image edge extraction,image segmentation,target recognition and multivariate filtering is realized.The main research work of the thesis:(1)Aiming at the complex correlation problem of remote sensing image,the fuzzy similarity measure is defined by the distance between vector and angle in RGB color space to describe the color similarity relationship consistent with human visual perception,and its characteristics are deeply studied and analyzed,and establish the fuzzy color similarity measure framework.(2)From the perspective of mathematical morphology,the method of changing the area-oriented is based on the fuzzy color similarity,and constructs the color morphology with pixel as the core,which can not only effectively "dilation" and "erosion" the object,but also the target has good deformation and smoothing ability.It effectively solves the image distortion caused by not considering the correlation of color components,and also avoids large deviation of operation results due to different sorting rules.(3)Based on the fuzzy color similarity,combined with color morphology,the image processing method suitable for color image is designed and developed to realize remote sensing image edge extraction,image segmentation and multivariate filtering.The thesis applies the above research results to the actual processing of high-resolution remote sensing images such as QuickBird,Ikonos and aerial.The results show that the proposed fuzzy color similarity measure framework,color morphology and image processing algorithms can meet the needs of actual image processing.It has a good application and promotion value,and also provides a research idea for the processing of remote sensing images.