| In recent years,air and sea accidents occur frequently,and many wrecks of aircrafts or ships have sunk to the sea floor.In order to salvage these sunken targets(such as aircraft black boxes,etc.),it is very important to search for these targets.For these artificial underwater target debris,the usual search method is to use Autonomous Underwater Vehicle(AUV)equipped with sonar and underwater low illumination camera to scan and image in a certain range of sea areas.In the search process,because of the large search range,underwater low-illumination cameras will take a very large number of images,which often contain a large number of underwater background images without artificial target debris,while the number of target images containing artificial target debris is very small.For these seabed images,manual interpretation is usually needed to find artificial target debris from the image,so a large number of seabed background images greatly increase the workload of manual interpretation.How to detect the human target debris in a large number of seabed images is the core issue of this paper.Based on the analysis of the complex background characteristics of seabed image and the characteristics of artificial target debris,a seabed image artificial target detection method is proposed,which combines edge detection algorithm,line detection algorithm and support vector machine(SVM)classification detection algorithm.According to the fact that seabed background image has more micro-edge information,this paper proposes an edge detection algorithm based on Canny operator and micro-edge filtering.Firstly,median filtering is applied to seabed image to eliminate part of the noise.Secondly,the traditional Canny operator edge detection operation is applied to the image to get the edge image of seabed image.Finally,the edge image is filtered by micro-edge filtering to further reduce the noise interference in the image.According to the obvious linear characteristics and regular shape characteristics of man-made target debris,a probability Hough transform combined with gradient direction filtering algorithm is proposed in this paper.Firstly,the edge image obtained by the edge detection algorithm is processed by probabilistic Hough transform to detect whether there is a straight line edge in the image and filter the detected straight line in gradient direction.Then,according to the number of detected straight lines,the detected image is divided into target image containing artificial target debris,suspected target image containing suspected artificial target debris and non-artificial target debris.There are three types of background images of skeletons.For target image and suspected target image,the contour of their edge image is searched,and the target area and suspected target area are determined by the detected linear edge and contour information.In SVM classification and detection algorithm firstly,the seabed background image and artificial target debris image are used as training sets to train SVM.Then,the trained SVM model is used to detect the suspected target image in the linear detection algorithm,and the target image and background image are obtained.In order to verify the effectiveness of the proposed algorithm some seabed background images and artificial target debris images are randomly extracted to verify the proposed algorithm.Through the analysis of experimental data,it is proved that the proposed algorithm can detect the debris of man-made targets in seabed images,which greatly reduces the time required for subsequent manual interpretation. |
