Research On Edge Detection Method Of Seismic Image Based On Improved HED Network

Edge detection technology is a basic problem in digital image processing and computer vision.With the rapid development of deep learning,edge detection algorithms based on convolutional neural networks have become a research hotspot in image processing.After years of research,the edge detection algorithm on strong edge images has achieved good results,but on weak edge images,there are still some problems to be solved.Therefore,this paper studies the edge detection methods of seismic images based on the improved HED network.First,a brief introduction to the basic theoretical knowledge involved in this article.In-depth study of the HED network edge detection method based on deep learning,the principle introduction,experimental comparison and quantitative analysis are carried out,and the advantages and disadvantages of the HED network are summarized.Then,in view of the problems of incomplete and unsmooth edges in the edge detection of HED network,an edge detection algorithm based on improved HED network is proposed.On the basis of the original network,two pooling layers are reduced,the deconvolution layer in the last two side output layers is modified,the HED network model is optimized,the output accuracy of the side output layer is improved,and the loss function is optimized.Compared with the HED network,the F-measure is increased by 2.3%,and the experiment proves that the performance of the improved HED network is improved.Finally,the improved HED network is applied to the weak-edge image seismic image.Binarize the edge probability map output by the improved HED network to obtain the salient edge.Use the edge extraction method based on matched filtering to extract the edge of the image,and merge it with the edge extracted by the improved HED network to get the final result.This method can reduce non-target edges to a large extent,and can effectively extract complete and accurate target edges.