The Study On Segmentation Of Aurora Oval In Aurora Image

The aurora is commonly appeared in the high latitude regions of the north and south poles,which is a beautiful natural phenomenon.The study of the activities of the aurora is not only important for the understanding of space weather and solar storms,but also helps to further study the structure and energy storage of the earth's magnetosphere.Usually,the aurora oval is an important parameter in the study of aurora activity,so how to accurately and effectively extract the boundary of the aurora ovals is the key problem in this thesis.The thesis mainly analyzes and studies the segmentation method of the aurora oval in the aurora image.Firstly,the original aurora images need accomplish image preprocessing,which can reduce noise and enhance the image of the contrast.Secondly,using an improved region growth method to extract the auroral oval boundary,and the mathematical morphological operation is used to modify the results of segmentation.Finally,according to the segmentation evaluation method,the segmentation results of this thesis can be evaluated.The main research contents of this thesis are embodied in the following aspects:(1)Prior to the aurora image segmentation,image preprocessing operations need to carry out.According to the characteristics of aurora image,image preprocessing mainly includes the removal of background,image denoising,image enhancement.Through the image preprocessing,on the one hand it can effectively remove image noise and avoid the misclassification of regional growth in initial seed pixels.On the other hand,it can enhance the contrast of the original image,which is conducive to the selection of growth threshold.(2)In region growing process,according to the aurora image gray histogram,using Gaussian fitting method for selecting a set of pixel value as the initial seed growth point,so it is effectively overcomes the defects of single initial seed point,and avoids the phenomenon of regional false growth caused by noise interference.Then,based on the analysis of the pixel neighborhood gray characteristics,the best segmentation threshold is obtained by using Otsu,this method replaces the system errors which caused by the traditional region growing method.Finally,the segmentation results are adjusted and modified by mathematical morphological operations,which eliminates the holes and discontinuities caused by noise interference in the segmented regions.(3)Evaluating the performance of the segmentation results by using the difference experiment method.By calculating the four quantitative indicators of the segmentation results,this thesis evaluates the advantages and disadvantages of the segmentation method,and a variety of methods are presented and compared with the experimental data of this thesis.According to the relevant experimental data,it can be concluded that the segmentation method used in this thesis has a high degree of accuracy.Under the Matlab7.0 platform,completing the aurora image experiment simulation.The results show that the segmentation method in this thesis can be more accurate and more complete segment the auroral oval region.At the same time,the method has good robustness,especially in the case of low contrast and high noise,it also can be more accurate to extract the aurora oval region.