Research On Underwater Laser Gated Imaging Target Recognition Technology

Seabed resource survey,marine biological monitoring,submarine search and rescue,and underwater archaeology are application areas where underwater optical imaging has significant scientific implications.Currently,gated imaging is the primary underwater optical imaging method.Although this method may eliminate environmental background noise and backward scattering noise outside of the chosen slices,it is still affected by the backward scattering noise inside the slices in turbid water,which reduces target recognition distance and recognition rate.An underwater polarization-gated imaging technique that combines optical polarization technology with gated imaging to solve the previously mentioned problems is proposed in this paper.A theoretical model of underwater polarization-gated imaging evaluation index（signal-to-noise ratio）was established and simulated.Several targets were identified,tested,and analyzed using a polarization-gated imaging experimental system that was developed.The findings of the experiment were evaluated and compared using convolutional neural network methods.The main research contents of this paper are as follows:（1）The intensities of signal beam,backscattered beam,and forward scattered beam of the target to be measured in the imaging system are calculated independently based on the principles of underwater gating imaging and polarization-gated imaging technology.A theoretical model whose evaluating indicator is the signal-to-noise ratio is constructed based on underwater polarization-gated imaging,and the model is then simulated and analyzed.The findings demonstrate that there exists a critical attenuation coefficient（c₀）when polarization-gated imaging is applied to underwater targets.The identification distance of polarization-gated imaging is larger than that of gated imaging when the water attenuation coefficient is greater than c₀.（2）Based on the"Phoenix Eye 3"underwater gating imaging system developed by the Beijing Institute of Semiconductors,Chinese Academy of Sciences,this study builds an underwater polarization-gated imaging experimental system with the inclusion of polarizers and analyzers,which can filter out the backscattering beam of the test target.Both conventional gating imaging and polarization-gated imaging were used to identify the target USAF1951 in the water tank environment,and the findings demonstrated that the polarization-gated imaging effect was greatly enhanced by introducing a polarizer and a detector as compared to conventional gating imaging and polarization-gated imaging with only one polarizer.Underwater gating imaging and underwater polarization-gated imaging experiments were carried out in a field pool at night with isolated ambient light for five targets:fishing nets,coral 1,coral 2,water plants,and target targets.The experimental results demonstrate the existence of a critical attenuation coefficient c₀.When the attenuation coefficient of the water exceeds c₀,polarization-gated imaging has a longer recognition distance than gating imaging,which confirms the conclusions of the simulation presented above.（3）Fishing nets,coral 1,coral 2,aquatic plants,and target imaging data sets from field pool experiments were collected,analyzed,and trained using a convolutional neural network based on the Res Net32 structure.The findings indicate that,under high turbidity,the target identification rate of the underwater polarization-gated imaging system may be improved by around 10%–20%when compared to the conventional underwater gating imaging system.This finding shows that the target identification rate may be significantly increased using polarization-gated imaging in water with a high attenuation coefficient.The target recognition rates of the neural network algorithm-based selective gating imaging and polarization-gated imaging system can be improved by about 20%and 30%–40%,respectively.The results show that the neural network algorithm can further improve the recognition accuracy of underwater target images in water with high attenuation coefficients.The findings of this study offer a novel approach to solving the target recognition issue in extremely turbid water.