Image Acquisition By Multiple Linear CCD And Application In Under Vehicle Defects Inspection System

Since9.11terrorist attacks in the United States, the threat of terrorism and publicsecurity issues focus the attention of all countries all over the world. Due to its goodconcealment and portability, many criminals use the vehicle to carry dangerous goodsand contraband, what does great harm to people’s lives and social stability. Therefore, todo under vehicle inspection is an essential project in the important places. Comparedwith the area array CCD camera, linear array CCD camera can scan the chassis rapidlyand continuously, adapting to a high speed without image distortion. So under vehicleimage acquisition of chassis under vehicle based on multi-linear CCD camera becomesa hot topic in the field of security.In the dissertation, its main content is about geometric correction of linear CCDcamera and image mosaic (contained image registration and image fusion) of multi-linear CCD camera. At first, research status about geometric correction and imagemosaic algorithm at home and abroad is studied and summarized. Then, workingprinciple of linear array camera is studied to make theoretical preparation for thefollowing study. On the basis of previous studies, the dissertation proposes a geometriccorrection algorithm combined with Surf and phase only correlation. First, thedissertation extracts feature points using Surf algorithm, constructs feature vectors,matches the feature points using Euclidean distance and estimates the homographybetween the two images using the relationship between the matched points. The methodin the dissertation was to use the homography to transform the distorted images and usethe phase only correlation to adjust the position in order to reach the aim of geometriccorrection. In the image mosaic, the dissertation proposed an automatic image mosaicalgorithm made of image registration based on Surf and image fusion of fading out torealize image mosaic.In the last part, the dissertation uses experimental platform to collect images as realunder vehicle inspection system. The experimental results show that compared with the traditional geometric correction algorithm, the method used in the dissertation combinedwith Surf and phase only correlation does not need to set control points and the imagequality was higher than the old methods. The Surf algorithm combined with fade outimage fusion algorithm is also well solution for the problem of blanket stitch, and theresults show that the method dissertation proposed can make the result realizesmoothing and seamless.