Research On Ship Detection Method Under The Background Of Sea And Sky

Target detection is an important research direction in the field of computer vision.It has a wide range of applications,such as text recognition,pipeline and robotics.For the unmanned ship autonomous navigation,marine environment monitoring,marine accident rescue and other fields,the detection of ship targets has important application background and practical significance.However,the maritime environment is complex and changeable.Ship motion,weather changes,illumination,sea surface reflection,ripples and other disturbances make image-based ship target detection face many problems.This paper demonstrates the detection method based on salient features and the method of region segmentation to realize the ship target detection under the seasky background environment.This paper focuses on the detection technology of ship targets in sea-sky background in visible light images,and adopts a detection method based on saliency features that fuse region segmentation.It is mainly divided into three parts: sea-sky background image enhancement,saliency map calculation,and target area segmentation.The specific work is as follows:(1)This paper analyze the detection results of 10 state-of-the-art salient detection methods in the sea-sky background.The analysis of the environmental characteristics of the sea-sky background shows the main factors affecting the salient detection results.So,we get the quality evaluation coefficient of the sea-sky background environment image: C,and then we analysis the method for optimizing the image quality of the sea-sky background environment according to the experiment.(2)In order to improve the result of detection.This paper fuses the information of sea-sky background.According to the characteristics of the selected salient detection algorithm,We analysis and verify the parameters that affect the detection result.Fuse the sea-sky background regional characteristics to optimize the algorithm selected in the second chapter.According to the experiment,We get the weighting parameter,and fuse the two algorithms that are improved to form the saliency map,and check it by experiment.(3)In this paper,We analyze the existing image segmentation algorithms,and fuse them with the saliency map,optimize the image segmentation results of the existing algorithms,and then I introduce local feature detection algorithm,optimize the post-processing results,and fuse it with the improved algorithm.At last,We check the result with experiment.