Research On Underwater Cable Recognition And Tracking Technology Based On Binocular Vision

In the process of social development,underwater cables have become one of the indispensable means of transportation for communication data and electric energy due to their economy,safety,strong reliability,and fast transmission speed.The use of underwater cables can greatly save energy costs and produce huge economic benefits.However,seawater corrosion,material fatigue,and changes in the earth’s crust will cause cable damage and fracture.Therefore,the regular cruise inspection of underwater cables has important research significance.The inspection technology based on machine vision is widely used in underwater cable inspection due to its intuitive and reliable images,non-contact and fast response.Aiming at the problems of poor real-time performance of traditional cable inspection systems and large computing resources required,a real-time cable inspection system is proposed,which takes real underwater cable images as the research object,adopts wavelet transform and improved γ-CLAHE fusion algorithm to complete image enhancement,Proposed an improved probability Hough transform and least square method to detect cable image sequence,to achieve accurate detection and continuous tracking of underwater cables.The main research contents are as follows:(1)Aiming at the problem that a single image enhancement algorithm cannot adapt to the changes in the underwater complex environment,and the darker background image is prone to over-enhancement and color cast,a wavelet transform and an improvedγ-CLAHE fusion algorithm are proposed,and the same The traditional CLAHE algorithm,Retinex algorithm,and Singh’s fusion algorithm are compared,and the effectiveness and superiority of the enhanced algorithm in this paper are verified through objective evaluation indicators.(2)Aiming at the shortcomings of traditional cable detection algorithms that focus on single-frame image processing and detection,the detection time of video sequence images is not significantly shortened,and an algorithm combining improved probability Hough transform and improved R-LSDR is proposed for video The image is detected,and the Kalman filtering algorithm is introduced to predict the direction of the cable,which further shortens the detection time and improves the tracking effect of the algorithm.(3)The imaging model of the binocular camera is established,the internal parameter matrix is obtained through the calibration experiment of the binocular camera,the conversion relationship between the pixel coordinate system and the world coordinate system is deduced,and the spatial position of the target relative to the camera is calculated.Through experiments,verification and evaluation to understand the accuracy of the calculation model.(4)An outdoor pool test scene was built,the proposed underwater cable detection system was tested,and the collected video was processed offline.The results show that the proposed detection system can accurately detect and track underwater cable images and meet the real-time requirements of the system.