Research On Visual Image Restoration Of Air Duct Robot

Center air conditioner has brought many advantages to human-being.Nonetheless, the attendant problems of a center air conditioner can not be ignored.The accumulating dust in the duct could be well breeding grounds for germs andbacteria which could accelerate the rate epidemic prevalence. While it is a pressingproblem to remove dust from inner duct, the limited space inside the duct makesmanual cleaning impossible. Therefore, an air duct robot as an assistant cleaningmachine is developed. The core of developing a duct robot are the research forautonomous navigation which highly relies on environment information. Althoughimages contain the most information of environment, images got in practice are oftendegraded due to the limitation of the working space. As a result, restoring degradedimages as good as possible are the precondition for effective acquisition of theenvironment information. Main problems are blur caused by relatively motion and fogformed with dust.For the research of dealing with the motion blur, an adaptive variational partialdifferential equation (PDE) based approach for restoration of gray level imagesdegraded by a known shift-invariant blur function is presented using Hopfield neuralnetwork. In the proposed image restoration model, an adaptive regularizationparameter is developed instead of the constant regularization parameter used inprevious PDE model. The value of the adaptive regularization parameter changesaccording to different regions of the image to remove noises and preserve edge better.Several computer simulation examples involving synthetic and real images are givento illustrate the usefulness of our method. Comparisons to the previous PDE basedmethod are also given. results show that the image restoration results of the proposedmodel both look better and have better SNR (Signal to Noise Ratio) than the previousvariational PDE based model and traditional model.Regarding the research of the restoration of the haze images, a window basedapproach are proposed to estimate the airlight inside the duct. With the methodproposed in dark channel prior image dehazing algorithm, the processed images havewell-proportioned gray level histograms in each RGB channels, and with a better meanvalue, standard variation and sharpness, respectively.