Bank Notes Fragments Reassembly Method Based On Digital Image Processing

In our daily life, it is often required to reassemble fragments for recovering the original shape. For example, the judicial department need to reassemble broken files, fingerprints, footprints and so on to extract evidence. In the field of archaeology, unearthed fragments need to be recovered. The research of banknotes fragments reassembly is also very useful in real life. Occassions that a pile of notes change into fragments due to human or non-human factors often happen. According to regulations, notes can be exchanged if the area of fragment is at least half or more of the entire area. Hence, these broken notes need to be reassembled in advance. Because of the limits of manual, such as time-consuming manpower and material shortage, it is difficult to complete. Therefore, it is necessary to search for a method of programming for the fragment reassembly.The thesis presents several measures of stitching fragments, as following:(1) Because of the particularity of banknotes:In the same set of RMB, the banknotes of different denominations have different colors. So the color histograms of the fragments images are calculated according those color features. According to the statistical results, the fragments of different denominations can be classified.(2) Since the whole image of different denominations is known, image matching can made by using the whole banknote as a reference image after judging which denominations the fragments image belong to. The method used in this thesis is the normalized gray correlation algorithm. When the image matching is performed, the angle of rotation of the fragments relative to the whole banknote will be calculated. This thesis adopts the gray correlation and phase correlation algorithm.(3) Once the effect of extracting color information is not good or the image matching is unsuccessful, it needs to extract its outline features. The polygon approximation can be used to represent the outline of fragment image.Through a series of strings, Freeman chain code can also be used to represent outline fragments of the image. According to the outline feature, the final reassembly can be made, and then judging the goodness of fit of its edge.(4) When reassembling the fragments, a few of actual fragments are reassembled firstly, then the spliced effect is analyzed. In order to make a joint study of all the denominations, the contour curves of the fragments are simulated and then a large amount of simulated fragments images are obtained. Finally through reassembling the simulated fragment images, some good results are made.