Method Study On Edge Detection Of Water Surface Target

As an important direction of computer vision,target edge detection is an important basis for subsequent target segmentation and recognition.In recent years,with the rapid development of video technology,it has been widely used in banks,hospitals,supermarkets,transportation and other public places.Our country has a vast territory,a long coastline and a wide sea area.The detection of water surface target has important practical significance and application value.It has played a very good role in military and civil aspects such as maritime rights protection,maritime supervision and marine environment protection.Because of the complex and changeable environment of water scene,the detection of water target is greatly challenged.In this paper,based on a large number of previous literature reading and previous research on edge detection,the main work is as follows:This paper introduces the basic concept and algorithm flow of Canny operator edge detection algorithm and mathematical morphology,and analyzes two problems in image edge detection using Canny operator.One is that it is not easy to set Gaussian filter parameters,and it lacks self adaptability to different images.Second,it is greatly affected by noise.In the first-order partial derivative finite-difference gradient calculation,because derivative calculation is very sensitive to noise,edge detection of image from two directions is prone to false edge and missing detection.Therefore,this paper improves the traditional Canny operator edge detection algorithm,using an improved bilateral filtering instead of Gaussian filtering,and use Sobel operator to calculate the gradient amplitude direction.In the process of gradient non maximum suppression,we use the maximum point as the candidate edge point,brush the maximum twice,and retain the non maximum points that are not adjacent to the candidate edge points in the 8 neighborhood.Finally,we use Otsu to calculate the image high and low thresholds.The results show that this method can More accurate detection of edge information,and has a strong adaptive.Secondly,the traditional morphological algorithm is used to detect the edge of the image.If the noise concentration is high,the detection effect is not ideal.Therefore,this paper proposes an improved morphological edge detection algorithm,which uses two different structural elements for edge detection.The improved anti noise morphological edge detection algorithm can reduce the noise very well,and the obtained image is relatively smooth,the contour is clear,and the information contained is rich.At the same time,the subjective effect and objective evaluation are improved.Finally,according to some difficulties in edge detection of water surface target,this paper adopts an edge detection algorithm based on wavelet transform to detect the edge of water surface target image.The edge image extracted by improved Canny operator edge detection algorithm and improved anti noise morphological edge detection algorithm is fused by wavelet transform to get the final edge of water surface target Edge image.The obtained edge image is better than the two improved methods in information entropy,root mean square error,correlation coefficient and distortion degree,and the visual effect is also satisfactory.