Research On Underwater Image Enhanced Visual 3D Reconstruction Method Based On Bionic Robotic Fish

Autonomous navigation is one of the cornerstones to realize the further development of intelligence and informatization for underwater robot.It is of great significance for accelerating the intelligent and information process of underwater robot and tackling key problems in the development of marine resources.In order to realize the underwater path planning and autonomous navigation functions of the bionic robotic fish,on the one hand,the robotic fish must be able to perceive environmental elements and utilize the known environment to guide itself to complete tasks,so as to realize the construction of environmental maps and its own positioning;on the other hand,due to the complexity of the underwater environment,the quality of underwater images is low,which is not conducive to the perception of the underwater visual environment.Therefore,in view of the low quality of underwater images caused by the complex underwater environment and the autonomous positioning of the underwater bionic robotic fish,this thesis combines underwater image processing technology and visual SLAM technology to propose an underwater image enhanced visual 3D reconstruction method based on bionic robotic fish and carry out experimental analyses.The main research contents of this thesis are discussed as follows:(1)A design method for the experimental platform of the bionic robotic fish is proposed.Based on the deduced and calculated physical model and kinematic model of the bionic robotic fish,the structural design of the bionic robotic fish is carried out,and the hardware system of the experimental platform is designed and implemented.At the same time,the software system of the experimental platform is described,and the overall design of the experimental platform is completed.(2)Aiming at the problem of low underwater image quality,an underwater image enhancement algorithm CB-GBCP based on the fusion of Color Balance and G-B Channel Prior is proposed.Based on the classic underwater imaging model,the RGB space color balance,the GB channel prior image restoration and the Lab space pixel adaptive stretching of the underwater image are completed successively,and comparative experiments are conducted with other algorithms to verify the applicability and superiority of the proposed algorithm from subjective effect and objective quality.(3)Aiming at the autonomous positioning of the underwater bionic robotic fish,an underwater visual 3D reconstruction method based on SLAM is proposed.Based on the calculation process of the classic visual SLAM system,the proposed fusion of Color Balance and G-B Channel Prior algorithm is used to optimize the key frame tracking.The feasibility of the proposed underwater visual 3D reconstruction method is verified by TUM standard data set and test data set.(4)Based on the underwater image enhanced visual 3D reconstruction experiment scheme of bionic robotic fish,underwater image processing experiments in different water environments,visual 3D reconstruction experiments in outdoor underwater environments,and motion trajectory tracking experiments are carried out.The results show that the CB-GBCP algorithm can effectively improve the comprehensive quality of underwater images,and can improve the feature matching efficiency,and then increase the map points by about 16.03%;the underwater map reconstruction results of the proposed method can describe the underwater environment well,and the average error between the real trajectory and the estimated trajectory is about 10 mm.These results verify the feasibility and effectiveness of the proposed method in realizing underwater map mapping and autonomous positioning of robotic fish.