Research On Algorithm Of Wreckage Detection For Deep Seafloor

In recent years,the plane crash occurred frequently,the plane crash occurred on the sea different from the crash occurred on land.The fuselage will fall into pieces and sank to the seabed,when the plane crashed on the sea.The detection of wreckage in the deep sea has great significance to salvage the black box.It's hard to salvage the wreckage in the deep ocean.When we search the wreckage is usually required AUV equipped with sonar or underwater camera,we using underwater camera images,and the target confirmed is by the human in finally.Because of the long working time of AUV,the number of images taken by underwater camera is very large,but the number of images containing the wreckage is very small.How to filter a large number of invalid images and improve the efficiency of detection is the core issue of this paper.In this paper,we studied the background of the deep see environment and the characteristics of wreckage,and we put forward the method about the detection of aircraft wreckage.This method was used to detect the image within the suspected target area,then to judge the suspected area.In suspected target area detection,first of all,because of the wreckage has obvious shape and line features,we used Hough transform line detection algorithm to detect the straight line in the image,and then marked the detection results in the image to enhance the effective edge.Then used Graph Based Visual Saliency algorithm to obtain the saliency map of the image,the region with the highest degree of significance was labeled as the suspected region.The support vector machine classifier was used to confirm whether the suspected target area is the wreckage.According to the characteristics of deep sea bottom image and the characteristics of the wreckage,we proposed four indices of average luminance,contrast,edge density and texture variance as the feature vector of support vector machine classifier.The training image database was composed of the deep sea background image and the aircraft wreckage image,then we used the database trained support vector machine classifier,which can be used to judge the suspected region of the target.In order to validate the algorithm,the experiments of deep water pool imaging and offshore seabed imaging were carried out,and we tried our best to simulate the characteristics and working environment of the deep sea.We obtained the data of the deepsea seabed simulation image,and did the target detection.Leakage rate and the ratio of the effective image in the total image were calculated.The experimental results showed that the proposed algorithm has the characteristics of low leakage alarm.When the effective image was reserved,a large number of invalid images can be filtered to reduce the number of images that need to be manually interpreted.