The Research Of Image Restoration In Dust Environment

Image restoration in dust environment is a creative work, and so far little research in this area has been discovered, but the problem is of its research significance. The research fruit can be widely adopted in the fields of rescue, traffic, production safety, agriculture and military. And the research on image restoration in dust environment in the dissertation has been carried out as follows:The dust as a kind of common material was summarized. Its definition and classification have been stated, as well as the physical characters. With examples, the dust on image brightness, contrast and sharpness are also stated here. Currently, due to fog, haze, clear image of the environment results in more, the author gives the overview and analysis on current status of the research, in order to learn from it and carry on the research on image restoration in dust environment. Finally, a method of research on image restoration in dust environment formed.Light scattering and absorption by dust particles is main reason for image degradation. So a research on the optical characters of a single dust particle was earried on in the dissertation. First of all, a theoretical analysis on light scattering intensity scattered by single dust particle, its propagation direction and energy absorption by dust particle was given. Then, the progress was simulated by Mie scattering programe. Finally, the optical characters of single dust particle were obtained from simulated results. The research provides the optical basis for establishing the image degradation model in the dust environment.On the basis of optical characters of single dust particle, the research on light transfer in the dust medium was discussed. First, two hypothesis were made for the dust medium, one was the hypothesis of independent and the other was the hypothesis of homogeneity. Second, the multi-scattering theory was used to give a theoretical analysis on the process of light transfer through the dust, and the way of luminous intensity variety in the light transfer process was discovered, as well as to image formation. Then, the image degradation model in dust environment based on the first order multi-scattering model was proposed. Finally, the image degradation model was used to synthesis the image, which provides experimental source material for image restoration in dust environment.According to the image degradation model, an algorithm of image restoration in dust environment was studied. First, the principle of the dark channel prior was introduced, meanwhile, its existence extensively in the nature was also proved with the experiment. Secondly, on the basis of image degradation model in dust environment, the dark channel prior was applied and single image restoration equation was deduced as well as depth map. And then, we explained the influence of each parameter in the image resumes formulae upon the result of image resumes influence through the experiment. Then, according to image objective quality assessment theory, a new image quality assessment function based on Threshold-Kirsch operator was proposed; it was used as a criterian for genetic algorithm to evaluate the parameters.then an optimal image restoration result was achieved. Finally, the algorithm was proved to be effective on dedust, and was able to recover detailed edge of the image. The restoration result provides basis for automatic objects recognition.Besides, we also made a research of the application of image restoration in dust environment. As we know, optical sensors on the rescue robot are often disturbed by dust, and this causes a few troubles for the obstacles detection and recognition. In order to solve this problem, a method of obstacle detection and recognition which is based on camera geometry model and the depth chart in dust environment was adapted and it could detect obstacles of arbitrary shape at any time with only a single camera and a single image. The robot's three dimensional coordinates can be obtained via the method, too. Therefore, based on it, in the way of the depth edge detected we were able to segregate the obstacles from the background. Finally, the validity and feasibility of the method was fully demonstrated by the experiments. The method provides the rescue robot a simple and economical way to detect obstacles in dust environment.