Research On The Method Of Strip Multiple Source Image Fusion

Strip steel is widely used in the aerospace, shipbuilding, automobiles and home electrical appliances, and its surface quality directly affect the quality and performance of the final products. So the strip surface detection system is researched widely. Followed with the development of detecting technology, strip surface quality detection system with one single image acquisition method cannot meet the need of the users. Two or more acquisition methods are being introduced to the strip inspection system. It raises a problem of how to integrate the imformation of these acquisition systems. At the same time, the image fusion technology has been widely used to synthesize image information of many collection systems, and good results are achieved. Inspired by this, image fusion technology is introduced to the strip surface detection system.The paper gives a preliminary conception of implementation process of multi-source image fusion. Meanwhile, in the existied laboratory condition, the effective image fusion method of two groups of images acquired by matrix CCD under the different angle is studied. The paper has made beneficial research and improvement in the following aspects:1. Overall grayscale value of the source images acquired by matrix CCD under the different angle are different and the difference will affect the quality of the source image fusion. This paper presents the corresponding solution to this problem, and the experiment result support this solution.2. Image fusion algorithm are studied. Bandelet transform and pulse coupled neural network (PCNN) are studied. Bandelet transform used as a kind of multi-scale decomposition tool and the improved model PCNN used as a fusion rule, the Bandelet-PCNN image fusion algorithm is presented in the paper. Compared with some useful image fusion algorithms, at the same time, through the quality of visual and objective evaluation parameters to evaluate the fused image, the method proves to be useful. Experiments also present that fused image integrated the information of defects of two sourse images effectively. Experiments show that image fusion effectively integrated defects’ information of many source images. The fused image contains tiny and more comprehensive information of the defects.