Research On Laser Image Edge Detection Technology Based On Deep Learning

Visual information occupies most of the perceptual information acquired by human beings.With the development of scientific information technology,image processing and computer vision have gradually become an important research direction.The edge is a part of the local mutation in the image,which has strong discontinuity,but also contains important information such as the boundary of the image.The edge feature can be extracted from the edge,which is the edge detection.Edge detection is the first step in image processing,computer vision,semantic segmentation and other technologies.With the development of electronic technology,the edge detection technology began to rise.There were Roberts operator,Sobel operator,Prewitt operator,Laplacian of Gassian operator,and Canny operator.Canny operator is a milestone of the classical differential operator edge detection technology.In recent years,deep learning has been greatly developed,and it has become a technological revolution.Various deep neural networks continue to appear,such as the Convolutional Neural Networks(CNN),the Deep Belief Networks(DBN),the Stack Auto-Encode(SAE).Therefore,deep learning and deep neural network applied to edge detection become a new trend.Because of its good anti-jamming,energy concentration and linear propagation,laser images are widely used.As for the edge detection of laser images,in general,traditional edge detection methods are adopted.However,it is necessary to introduce a new edge detection method because of losing the real edge,detecting the false edge and inaccurate location,which caused by the noise and other problems.With the development of deep learning edge detection technology,it has been a new research topic to apply deep learning edge detection technology in the field of laser image edge detection in order to achieve better detection results.In this paper,the traditional differential operator edge detection method,deep learning and deep neural network theory have been studied systematically and deeply.Based on that,the advantages and disadvantages of edge detection operators are analyzed,and the edge detection technology based on deep learning is introduced.Then,we analyze several deep learning detection algorithms,including the Holistically-Nested Edge Detection(HED).After the analysis and comparison,the HED algorithm is mainly researched.We use the laser image as the image to be detected.We use MATLAB to implement all kinds of differential operators and HED algorithm,analyze and compare the detection results,and discuss the advantages and disadvantages of deep learning applied to laser image edge detection.Experimental results show that the deep learning edge detection technology represented by HED algorithm has better effect in detection continuity and detection accuracy in the field of laser image edge detection,compared with traditional differential operator methods.