Any Background Of The Next Class Of Rectangular Target Recognition Technology Research,

The thesis is a technical study of a class of rectangular target recognition on the arbitrary background, and is directed at color duplication shelter-wall technique, that is rectangular money recognition on the arbitrary background, of FC141 type multi-function office machine. Because RMB is typical rectangular target, it is general to be chosen as study object, what's more pattern of RMB is complicated, it is representative to be study.The original input image of the system is scanned by scanner, resolving power 300dpi; the system applies image compression technique, removing image noise technique, image segment technique, mathematical morphology, image recognition and space geometric transform technique to process and analysis, then proceed to recognize; at last it exports the result in image. There are two possible directions: (1) if there isn't 50 yuan or 100 yuan RMB pattern, the system outputs the original input image; (2) there is 50 yuan or 100 yuan RMB pattern, the rectangular region that is covered by RMB is color-to-gray process and the other region is exported unchanged.The thesis depends on the digital image process technique, applies mathematic morphology and image recognition to resolve problems. For example, the system use 4×4 image sampling technique to reduce processing data; smoothing filter to restrain image scanning noise; to separate characteristic object, employ inter-class variance threshold segment, in addition, in the course of picking up character dots, uses HMT and so on.The key technique of the thesis is characteristic selection and collection. The thesis selects Braille, character "0" and "G" that represent RMB, as character object. To Braille collection, HMT serra putted forward is employed. The method can detect object and its background at the same time, and protect collection accuracy. To character "0" and "G"collection, the thesis adopts improving transform method based on HMT, the new method is called variant H1%(1* (VHMT) in this paper. It is suitable for detecting holiow object, for instance "O" and "G", its character is that it can detect regions surrounded by object and peripheries of object at the same time. The VHMT doesn*s detect object itself outwardly and detects its surrounding and peripheral regions, but inwardly reflects geometric composition character of object by detecting the background regions of it, moreover the detecting speed is fastandtheefihctis ideal.The result of the thesis resolves a difficulty that FC141 type multi-fanction office machine prohibits money-digital color duplication. It can recognize precisely such as 50 yuan and 100 yuan RMB pattern; regardless of its face or back~ and straight or rotated. Moreover, if there are ninny RMB pattern in the image including covering each other, so long as their character objects exist completely, the system can recognize precisely.