Research On Clearness Technique Of Single UAV Image In Fog Weather

As is known that traditional Remote sensing and aerial photogrammetry need high costing, strict flight condition, and long mapping period, so it is hard to meet requirements that emergency security. Unmanned Aerial Vehicle Remote Sensing(UAV)with its flexibility, economical efficiency, adaptability, and rapid map production, which has played an important role in the field of land dynamic management, resource exploration, disaster monitoring and emergency measure. With the rapid development of UAV technology, the issue concerning effective utilization has attracted more attention.However, as the result of the atmospheric scattering and absorption, the UAV image shows the character of low contract, not much clear, blurring of details, false color and so on, which has seriously influent the accuracy and precision of later work to a large extent. The degradation image could be more useful by means of haze removal.Therefore to remove the haze of image has the far reaching significance. What’more, it’s difficult to recover real scene from single haze image. As the process of hazy UAV image clearness needs timing and high efficiency, which requires high demanding.This paper dig a little deeper into the basic theory and key technology of haze removal from the model based and non-model based perspective, besides, realizing and comparing the existing defogging methods, what’s more important is to explore some new ideas of removing the fog of UAV image effectively and automatically and timely.The specific research of this dissertation has done are listed as followings:Firstly, researched the typical defog methods, compared some kinds of key technologies, and analyzed their basic characters and suitability, also aimed at the feature of UAV image that the contrast tend to a kind of hue and depth of field changed little doing experiment with some of those methods, summarized their advantages and disadvantages.Secondly, emphatically studied the Retinex method from the perspective of image non-modeled enhancement, analyzed the progress and applicability of the three kinds of Retinex algorithms: SSR, MSR, MSRCR respectively, also making the use of those to remove the fog of remote sense image, finding out the appropriate parameters in the process of algorithms to achieve the preferable results, at last evaluated the results from subject feeling and object parameters counting.Thirdly, researched the haze removal method of using dark-channel prior which is based on the physical model, besides, optimizing and simplifying this method for thatthe heavy computation, complexity, and the limit of restoration based on the feature of UAV image. After that proposing an optimized method of dark-channel prior to realize defogging fast and effectively. The results through MATLAB shows that the optimized method has the advantages in effectiveness and real-time processing.Finally, on one hand to analyze and compare the processed haze-image from the significance of fog removal and restoration of details and the color fidelity, which is object evaluating, on another hand to evaluate the defogged algorithms from some objective image quality parameters such as the average gradient, contrast, and entropy..