Research On Vision System Designing And Image Dehazing Algorithm Of Underwater Vehicles

In recent years,underwater vehicles have been extensively used in scientific research and commercial activities,and the diversified needs in different applications have promoted the research of underwater vehicles in return.However,owing to light propagating in different kinds of optical media,the underwater image usually has different degrees of nonlinear distortion.Due to the inherent characteristics of the water medium,the light propagating in the water will be absorbed due to exponential attenuation and the scattering phenomenon,underwater images usually suffer severe low contrast and color distortion.Therefore,it is necessary to solve the problem of degradation and refraction of underwater images in underwater vehicles vision applications.In order to tackle the problem of underwater image distortion caused by refraction phenomena and obtain the geometric information of an underwater scene,a geometrical imaging model of underwater camera considering the phenomenon of quadratic refraction is established.After establishing the relationship between three-dimensional points in the world coordinate with two-dimensional points in the image coordinate accordingly.The underwater camera model parameters are obtained by improved checkerboard calibration method.Meanwhile,in order to verify the accuracy of the underwater camera model parameters,this thesis adopts the method of underwater parallel plane measurement,and calculates the distance from the corresponding corner point to the main point in the air and in the water respectively to verifies the model parameters.The experiment shows that underwater camera model parameters are valid.In order to correct the underwater image,this thesis establishes the geometric mapping relationship between the object point in a fixed depth and the corresponding object in the anhydrous environment,and then completes the correction of the underwater image.The parallel plane measurement method is utilized to verify the correction results.The experimental results demonstrate that the error of the image correction method in this thesis is within the effective range.In order to improve the underwater scene image,a method based on the image fusion strategy is proposed.The color restoration version and the contrast enhancement version of the input underwater image are respectively obtained by the dark channel priori theory and the histogram equalization method,and those images becoming an image group.The image group is fused by a specific image assessment weight matrix,and the best part of each image in the image group is retained to produce a fused image with enhanced visual effect.The image entropy is selected as the quality evaluation index,and the optimal parameters are obtained automatically from the initial parameters of the underwater dark channel priori model by using the downhill simplex algorithm.We have carried out comparison experiments with state-of-the-art underwater image enhancement methods.The experimental results show that the proposed method can improve the resolution of underwater image and restore the color information of an underwater scene image.