Research On Ship Borne Electronic Image Stabilization Based On The Fast Detection Of Sea Horizon

With the rapid development of computer vision,the application of sea unmanned boats is also becoming more and more widespread.In the process of human-computer interaction,the video camera shakes violently with the sloshing of the ship.It also leads to human eyes discomfort and even affects the control and avoidance of unmanned boats.Electronic image stabilization came into being,aiming to solve the problems of sharp shaking and vibration of video images and provide strong protection for friendly human-computer interaction.However,there is not a universal algorithm that can solve the motion of video images in all scenes.The processing speed of video images also restricts the application of complex algorithms.Aiming at the problems of scene application and existing algorithms,this paper proposes an improved algorithm to detect sea-sky feature points and fit the sea-sky lines.Then,according to the time-consuming analysis,three algorithms are proposed to enhance the real-time performance.And the experimental verification and the evaluation of the quality of the image were carried out respectively.The specific work is as follows:(1)The influence of the 6-DOF motion of the ship on the image motion is analyzed,and the conclusion that the pitch and roll influence the longitudinal translation and rotation of the image respectively is obtained.By analyzing the phenomenon of visual analogy,visual persistence and other characteristics of man-made evaluation to develop a stable image of the environment standards.And the shortcomings of the existing algorithms in this scene are analyzed,the error exceeds the standard and it is difficult to ensure real-time.(2)Analyze the reason why the sea-air line is non-linear in this environment and calculate its curvature in the image.In view of the characteristics of sea-air curve,the detection algorithm of sea-sky feature points based on the gray-bump characteristics is applied and improved,and the causes of failure in complex scenes are also analyzed.Then,the sea entropy feature point detection algorithm based on the characteristics of the proposed,according to the characteristics of line-average texture(GPA)texture,the feature points of the sea-sky are detected by calculating the variation of histogram entropy of grayscale difference.Then,the sea air curve is fitted by random sampling uniform algorithm(RANSAC).Analysis and improvement of uniform segmentation is uneven segmentation.Meanwhile,motionestimation curve is used to verify the adaptability and accuracy of the algorithm in complex scenes.The algorithm has better effect,but at the same time,it exposes the shortcomings of the algorithm of curve fitting,such as complex and long time consuming.(3)Analysis of improvement for the real-time performance of the system,put forward the resolvent set out curve fitting,using interpolation search algorithm to approximate the sea sky line;The detection area of the post frame sea air feature point is divided according to the motion law of the adjacent frame time.According to the consistency of the gradient direction of the image near the sea-sky feature points,a method to determine the confidence of sea-sky feature points detection is proposed.The above three improvements can not only reduce the calculation time,but also enhance the accuracy of the sea horizon detection.(4)In the Matlab platform,the motion compensation image compensation for the continuous frame image is compensated.Through the experimental analysis,the video has a better impression on the video after the image stabilization.In C++ environment call OpenCV library to deal with this video,the basic real-time electronic image stabilization,better than the traditional 5 kinds of algorithms.Compared with the traditional algorithms,the improved algorithm has better effect on the precision and the real-time performance of the complex scene,such as the non-straight line of the sea and sky.