RMB Fragments Reassembly Method Based On Digital Image Processing

It is an interest question to reassemble a pile of fragments. In our ordinary living, we are required to reassemble fragments into an origin figure in many occasions. For example, in the area of historical relics and archaeology, we need to recover the ancient potsherds we have unearthed. Another area is intelligence, there have chances to recover the paper fragments and then can find useful information in these recovered papers. Also there have similar tasks in criminal detection area, for example we need to recover the broken evidences such as fingerprint, we are demanded to reassemble the fingerprint fragments into a whole fingerprint. These tasks are still rely on manpower presently, they are tough and time-consuming. So, it is necessary to find a machine method to process these fragments.The problem of RMB fragments reassembly has origin of practical needs in living. When a large number of RMB notes changed into a pile of paper fragments because our poor storage, we would want to change these fragments into new notes from bank. But according to the relevant rule, these fragments can be changed only when they can be reassembled a note that its area is over 50% than a new note. This paper tries to develop a machine method to reassemble RMB fragments automatically.RMB fragments'reassembly has its own specific characteristics. First, its measurement is two dimensions comparing to potsherd, so its reassembly is relatively simple; second, comparing to ordinary paper reassembly RMB fragments should match not only with figure but also with the design on note. Because of the RMB fragments reassembly's specific characteristics, in this paper we complete reassembly task by following steps. First step is to find the fragments'correct position on the reference image. To find this position we usually adopt intensity-based correlation algorithm. But this algorithm has a precondition that the fragments should be parallel with the reference image. RMB fragments have a little chance to be parallel with the reference image, this paper adopts improved intensity-based correlation to solve this difficulty, at the same time we adopt optimal search to accelerate the searching velocity. Then first we can find the rotation angle between the fragments and the reference image, then rectify the fragment images'position by this angle, at last we use the ordinary intensity-based correlation algorithm to find the fragments'correct position on the reference image. When every fragments'position on the reference image can be find, we can continue the second step, judging what fragments are adjacent. This paper gives a new innovative way that skillfully uses image inflation and image corruption algorithm to solve the problem. This way has the advantage of fast calculating speed and good accuracy. All the steps above-mentioned have validated by experiments, and has been proven its feasibility.Having completed the key steps above-mentioned, this paper gives a GUI to help ordinary users can use this method conveniently. Users can push buttons on the GUI to accomplish the functions above-mentioned. Finally, some suggestions and expectations are put forward for further research.