Multi-scale Heliochromatic Fiber Identification Algorithm Based On Bilateral Filtering And Hessian Matrix

Fiber structure is a kind of linear elongated structure existing in the solar chromosphere layer,which presents jet state.Studying the dynamic characteristics of chromosphere fiber can help people to further study the evolution process of coronal mass ejection,coronal heating and needle.The premise of studying the dynamic characteristics of fiber structure is to accurately identify and analyze the solar chromosphere fiber Therefore,the selection of an efficient and accurate recognition method is the focus of current people's attention.Before the chromosphere fiber recognition,we first analyze the characteristics of the fiber image,and find that there are many difficulties and problems in the current fiber recognition.In the chromospheric fiber image,the intensity change of the fiber structure itself is not large enough,so the difference between the fiber structure and the background is relatively small,and the adjacent fibers are prone to cross and overlap.At the same time,there are still many other solar activities around the fiber activity,which will interfere with the extraction of fiber features.The above reasons affect the identification and extraction of fiber structure.In order to solve the above problems,this paper proposes a multi-scale solar chromosphere fiber identification method based on bilateral filter and Hessian matrix.Combined with bilateral filter and Hessian matrix,a multi-scale linear filter is constructed to identify and extract the fiber structure.We choose a group of fiber image data to explain the steps of the recognition algorithm in detail,as follows:(1)normalize the fiber image to reduce the influence of the subsequent transformation function on the image data,which is convenient for the subsequent function transformation;(2)remove the noise of the fiber image by using bilateral filtering technology,which can effectively remove the fiber;(3)the Hessian matrix eigenvalue and Gaussian function are combined to construct the corresponding multi-scale filter to identify the fiber in the filtered image,which can enhance the fiber structure in the image,solve the problem of unclear fiber edge,and better separate the adjacent fiber structure;(4)the next step is to identify the fiber structure The recognition image is binarized by selecting the appropriate threshold;(5)the binarized fiber recognition image is processed by morphology to remove the small area connected components,and the interference factors of non target features are removed,that is,other features around the fiber are removed;(6)the fine structure of the fiber can be obtained by morphological thinning,which is convenient for intuitive observation of the fiber structure Structure.To confirm the accuracy and applicability,we use the proposed method on the unused data.The final experimental results show that the recognition method proposed in this paper can accurately and effectively perform the recognition and extraction of the fiber structure in the graph,and can also better identify those fiber structures with weak strength.Finally,the threshold designed in the experiment is optimized The parameters and window size of the two-sided filter are discussed.The experimental results can better prove that this method has good results,accuracy and applicability in the identification of fiber features,and the extracted fiber structure can be used in the next step of scientific research.