Research On The Technology Of Registration And Mosaic For Defects Image Of Board Strip

At present, the strip manufacturers have an urgent need for steel strip surface defect detection, defect statistics and analysis, quality control, etc, in order to meet production needs and enhance their competitiveness. Image mosaic technology can remove the large amount of redundancy between adjacent image information, also can reduce the amount of data to a certain extent.Image registration can realize fast regis-tration of defect image, the result of mosaic can be used in image feature extraction and defect classification as the basis.Therefore it is very important to study on image registration and mosaic.Although the number of image registration algorithm is a lot, there is not any single method can be applied to every image.Most of the existing algorithms are widely used in medical and remote sensing applications.We have to find the balance among accuracy, speed, robustness in the algorithm,combining with the specific cir-cumstances of the steel strip image, and develope a new algorithm fit the steel strip image feature.The major works of this paper are listed as follows.Analysed the cold steel strip surface defect images, found out similarities and differences with remote sensing image or medicine, then sumed up the characteristics of steel strip image.According to the requirement of steel strip testing system characteristics of registration and mosaic can be summarized.The method based on wavelet transform of image registration was proposed, the defects hold smaller space with a concentrative distribution, these characteristics were taken fully into account by this method.Therefore choosen defect image block instead of the original registration can greatly reduce the computation time, also can improve the efficiency of registration.The image registration algorith based on small area and cross-correlation was proposed. For steel strip defect image contains linear defects with only, image matching to teatures region as the model image with phasestep linetype, can avoid misregistry caused by improper selection subblock used in wavelet transform method.It was optimized to achieve the seamless mosaicing on steel strip surface defect image.The registration method applied to image mosaicing and the seam was removed by using the method of linear gradient fusion.Experiment results show that, the algorithm in this paper applied to image regis-tration can improve calculative efficiency in their application fields.By realizing steel image registration in different types, it proved this method of image registration had practical value in its field.Using the method for linear gradient fusion to image mo-saicing achieved good effect, it proved this method had a promising application pros-pect in image mosaicing field.