Research On High-resolution Remote Sensing Image Segmentation Based On Region Merging

With the rapid development of satellite remote sensing technology,the spatial resolution of remote sensing images has been improved and its application has been more extensive.The extraction of typical ground object information from high-resolution remote sensing images is one of the important applications.High spatial resolution remote sensing images have the characteristics of abundant information and large amount of data,which brings new challenges to remote sensing data processing.Traditional information extraction of remote sensing based on pixel levels cannot meet the needs of analysis and application of high-resolution remote sensing images,and the object-oriented high-resolution remote sensing image analytic method has become an important research direction.Aiming at the characteristics of high-resolution remote sensing images,this paper generates the initial segmentation region by superpixel algorithm.By making full use of the detailed information such as spectrum and shapes,the initial segmentation region can be merged to achieve the object-oriented segmentation process.The main research work of this paper is as follows:(1)Taking the process of region merging algorithm as the main line,this paper summarizes the principle and algorithm of commonly used superpixel generation method,analyzes and compares the process and nature of the region merging models,it introduces the commonly used merging rules of region merging segmentation algorithm.(2)The basic principle of SLIC algorithm is analyzed,and the influence of SLIC parameters on segmentation results is analyzed through experiments.The segmentation results of superpixels generated by SLIC algorithm and other commonly used superpixel generation algorithms are compared,and the results are analyzed qualitatively and quantitatively.(3)Using spectral and shape heterogeneity criteria to calculate heterogeneity,SLIC superpixels are merged.In this paper,a SLIC superpixel multilevel merging segmentation algorithm based on spectral and shape heterogeneity criteria is proposed.By setting several sets of merge parameters to merge regions step by step,the merging process is more refined,and the computation of heterogeneity can be more targeted at the merging objects at the current level,and the effectiveness of the algorithm is verified by experiments.