Research On Edge Detection Method Of Remote Sensing Image Based On Adaptive Mathematical Morphology

As a simple and efficient image processing method,mathematical morphology has been applied to many aspects of image processing,including image filtering,image segmentation,edge extraction and target recognition.Mathematical morphology is a set theory-based method that uses structural elements as "probes" for image analysis and processing,and is widely used in binary and grayscale image processing.Classical mathematical morphology usually uses structural elements of the same size to process the whole image.Since different classes of pixels have different sensitivities to structural elements,the image will be over-processed or under-processed.In addition,because color images have multiple dimensional information,traditional mathematical morphology cannot be directly applied to color images.Consenquently,how to establish color mathematical morphology has become a hot area of research.Classical mathematical morphological operations can shift the boundaries of targets in an image,making the processed image produce false texture information,or causing the output image to lose parts of the image details because the targets in the image are too complex or diverse.To address this situation,the thesis proposes a method to compute anisotropic features of pixels based on neighborhood symmetry,and then constructs an adaptive structure element selection method.The proposed method first defines the neighborhood symmetry measure of the pixel,and expresses systematically whether the pixel is in a flat region or an excessive region,so that different types of pixel can be processed by different types of mathematical morphological operations.In addition,the thesis reshapes the traditional structural elements into two mutually independent structural elements,namely,the expansion structural element and the erosion structural element,with the aim of avoiding the intersection of the operational processes of expansion and erosion operations for certain pixel points.On this basis,the color adaptive mathematical morphology operations based on neighborhood symmetry,including expansion and erosion,open and closed operations,are redefined,the properties of the operations are discussed and analyzed,and a new adaptive mathematical morphology-based edge extraction method for high-resolution remote sensing images is proposed.In order to compare the color adaptive mathematical morphology proposed in the thesis with the traditional mathematical morphology from several perspectives,the thesis firstly makes a qualitative comparison in terms of image detail capturing ability,and introduces four image quality evaluation indexes for quantitative analysis,and finally conducts experiments and analysis on the proposed edge extraction method based on adaptive color mathematical morphology.The experimental results show that the mathematical morphological operators proposed in the thesis has better performance in image detail capture,texture preservation,and image distortion degree,and the proposed adaptive mathematical morphology-based edge extraction method can effectively identify various types of edges in high-resolution remote sensing images.The paper has 33 figures,5 tables and 66 references.