Method and arrangement for the recognition of characters and/or objects

摘要

For the recognition of unknown characters and/or objects, the characters and/or objects are first read optoelectronically and the reading signals are digitised to generate a binary one- or two-dimensional image of the characters and or objects, which is compared pixel-by-pixel with reference raster images of known reference characters and/or reference objects stored in reference stores for coincidence and/or deviation. In order to achieve a higher accuracy and to be able to process any arbitrary characters and objects, the external dimensions of the binary image of the character and/or object to be recognised are determined and compared with the external dimensions of the binary reference raster images of the known reference characters and/or objects. For the binary images to be compared in each case, a counting surface is determined whose size is a function of the respectively largest values of the external dimensions of the binary images. The larger binary image is displaced relative to the smaller binary image and relative to the counting surface, and the pixel-by-pixel coincidence or deviation of the binary images in the respective output position and in the mutually displaced positions inside the counting surface is determined. Subsequently, the maximum degree of coincidence of all the highest degrees of coincidence from the comparisons is determined, whose associated reference raster image most closely [lacuna] the character and/or object to be recognised ... Original abstract incomplete.