Research On Enhancement Method Of Intelligent Ship Vision System In Sea Fog Environment

In recent years,the degree of informationization and intelligence of ships has been continuously improved,and it is currently developing in the direction of unmanned and even anthropomorphic.As the core technology of the intelligent navigation module,the ability of the intelligent ship to perceive the surrounding environment is an important prerequisite for ensuring that the ship can work normally in a strange environment and realize its function.The intelligent ship collects surrounding environment information and performs online analysis and feedback through various instruments and equipment.Among them,the visual system is the key to environmental perception,which is mainly composed of one to more optical cameras.However,the marine environment is complex and changeable,and often accompanied by sea fog.Under the influence of sea fog,the fine particles suspended in the air will scatter the incident light,and the image captured by the camera will be seriously degraded in visibility and contrast,resulting in a significant reduction in the accuracy and reliability of the information obtained by the visual system,seriously affecting the performance of the intelligent ship vision system,resulting in the loss of tracking targets or even due to the inability to find ships or other obstacles on the navigation path in time and failing to make effective avoidance measures resulting in collision accidents,which directly endangered the intelligent ship’s navigation safety.Therefore,only by improving the clarity of the images collected by the vision system in a foggy environment can the intelligent ship have better environmental adaptability and continue to complete its tasks and missions under the influence of harsh weather.The research in this dissertation is mainly based on the visual system enhancement technology under sea fog:Firstly,in view of the problems of poor and unsuitable defogging caused by using most of the current land defogging algorithms to defog sea fog images,a method for defogging sea fog images on the background of sea and sky is proposed.Inspired by the dark channel prior theory,the concept of bright channel is introduced,and the two methods of transmittance estimation are combined by combining weights,and the sky area prone to color distortion after defogging is appropriately corrected,Finally,the improved atmospheric light value and the corrected transmittance are substituted into the atmospheric scattering physical model to obtain a clear image,and simulation experiments verify the effectiveness and superiority of the algorithm.Then,in view of the shortcomings of traditional image defogging methods that require manual adjustment of parameters to obtain a better defogging effect,a generative adversarial network model considering channel and pixel feature information is designed.By using the Wasserstein distance as a trade-off measure for the discriminator network,and adding a residual module based on the attention mechanism to the model structure of the generator to improve the quality of the generated image,it is finally obtained during the game between the generator and the discriminator a defogging image with better visual effects.The results of this model are compared with the defogging images of several other models subjectively and objectively,which verifies that the model has better defogging ability.Finally,according to the characteristics of intelligent ship navigation,a fast video defogging method combining wavelet transform and guided filtering is proposed.When the ship is in a stable navigation state,considering the transmission of fog features between frames,the guided filter is used to smooth the fog feature map to retain the maximum detail information,and the subsequent video frames are directly defogged through the fog feature map;when the ship motion state changes greatly,the fog image is decomposed into four parts by wavelet transform and only the low frequency component is used to defog a single image.The experimental results show that the method can effectively improve the efficiency of defogging while ensuring the effect of defogging.