Research On Image Preprocessing Of Large Format CCD Digital Aerial Camera

With the development of micro-electronic technology and the advancement of surveyingand mapping technology, large format CCD sensors instead of traditional aerographic film anddigital images acquired by aerial camera have been a main tendency in the fields ofphotogrammetry and remote sensing. As a key step in image processing, digital aerial imagepreprocessing technique has become a hot topic in recent years. The thesis focuses on imagepreprocessing of large frame aerial digital camera DMZ, which is a state-of-the-art cameradeveloped by China independently and is designed by the idea of the ‘6+1’ CCD compoundcamera. The key techniques of geometric rectification and radiometric correction of DMZ’sprimitive panchromatic images are discussed and analyzed. A new method of geometricrectification is proposed, which six sub-pan images are merged to a full frame image. Templateof radiometric correction of the DMZ camera is designed through uniform illumination test inlaboratory. As an example of images acquired by DMZ camera in Hanzhong’s calibration site,images processing workflow and valid solution of the virtual image is designed, tested andverified. Furthermore, based on the typical haze phenomenon in aerial digital images, animproved haze removal method is provided. A set of experiments and analysis have been made.The main works of the paper are as follows:1. The development of large frame CCD aerial digital camera is reviewed. The performanceof several representative large frame aerial digital cameras is introduced at home and abroadnowadays. Current image preprocessing technology schemes of different cameras are analyzed.Necessity and practical significance on studying image preprocessing are pointed out.2. Fundamental components of large frame CCD digital aerial camera DMZ and its imageprinciple are expounded. Factors which cause geometric and radiometric errors in panchromaticimages taken by DMZ are intensively analyzed. According to its unique light path segmentationdesign and error sources, error models are summarized for improving the image quality andaccuracy.3. According to characteristics of DMZ’s original sub-panchromatic images, the idea andmethod of radiometric correction are summarized to correct radiometric errors in raw sub-images.The method of radiometric correction based template acquired by laboratory illuminance tests isdesigned. The sub-images are processed and analyzed by applying the template to eliminateradiometric error. The experiment results are shown.4. An improved haze removal method of digital aerial image is proposed based on hazeimaging model and dark channel prior. A simple and convenient method is drawn to calculatetransmission in haze imaging model and, moreover, contrast adaptive adjustment algorithm for the haze-removal images is introduced. As result the experiment shows that this method is rapidand effective in removaling haze of digital aerial images.5. The mathematic models of geometric correction and the common resample methods arepresented. A new method of virtual image stitching and mosaic is put forward and designedusing multi-camera models. In order to validate the experiment results, aerial triangulation iscalculated by stereo virtual images to estimate the accuracy and quality. This is an exploratorystudy to new typical digital camera DMZ.