Research On Detection Method Of Underwater Moving Target Based On Polarized Optical Flow

Underwater optical imaging technology is widely used in underwater target detection and tracking,seabed mineral exploration,marine environment monitoring,and other fields.However,during the transmission of light in water,the particles in it will absorb and scatter,resulting in many problems such as image noise,low contrast,and blurred details.Therefore,improving the quality of underwater imaging,improving the ability of underwater target detection,and realizing underwater moving target detection have high research value and broad application prospects.Compared with traditional light intensity imaging,polarization imaging technology can detect multi-dimensional information of target reflected light,characterize target property parameters that are difficult to distinguish by light intensity and improve the detection ability under low contrast conditions.Based on polarization imaging technology,this thesis conducts research on image de-scattering in turbid water and moving target detection based on optical flow.Through the analysis of current de-scattering methods and optical flow algorithms,a simple and effective polarization de-scattering method is proposed.Based on the de-scattering method,a polarization optical flow estimation method is proposed to realize the detection of underwater moving targets.This paper mainly conducts the following research:(1)Given the problems of existing algorithms such as accurate estimation of priori conditions and insufficient scattering removal in turbid underwater environments,an adaptive de-scattering method based on specific polarization states is proposed.Reflected light and scattered light have different polarization states.The optimized specific polarization state interval is used for polarization filtering to suppress the scattering of water particles.Through the imaging experiments in turbid water,compared with other de-scattering methods,the results show that,compared with the original images of degree of polarization and angle of polarization,the objective evaluation index EME value has increased by 28 % and 455 %,the variance has increased by 38 % and 124 % and the average gradient is improved by 6 % and19 %.Therefore,this method can effectively overcome the influence of scattering and improve the image contrast.(2)Aiming at the problems of non-conservation of brightness caused by strong underwater scattering and the inability to accurately estimate traditional optical flow,a polarized optical flow estimation method based on LK optical flow is proposed.By introducing the polarized optical flow constraint equation,the influence of underwater scattering on the optical flow estimation is removed.The effectiveness of the polarized optical flow estimation method is verified by the underwater moving target detection experiments with different turbidity,the results show that in a turbid underwater environment with 1.5 ml of milk solution added,compared with the scattering medium optical flow method,for metal targets,the average angle error is reduced by 63.08 % and the average end point error is reduced by 80.96 %;for plastic targets,the average angle is reduced by 28.18 % and the mean endpoint error is reduced by46.99 %.Finally,the moving target is segmented on the polarized optical flow field and the binary image of the underwater moving target is obtained.