GB1359782A - Character location and deskewing system for character recognition systems - Google Patents

Abstract

1359782 Character recognition BENDIX CORP 3 Jan 1972 [7 Jan 1971] 158/72 Heading G4R Characters to be recognized are accompanied by locating marks which are used in the determination of the location and skew of the characters within the field of scan. The scanner scans across the field in a succession of parallel lines. If a locating mark is detected in a preset number of successive scan lines the output of the scanner during the following scans is passed to a processor to relocate, deskew, and recognize, the characters. As disclosed the field of scan 14, Fig. 4, encompasses three letters, the first and third characters being aligned with locating marks 16, 17 as shown. The scan lines are horizontal, the first scan being at the bottom of the field, and scanning is initiated by a sensor which detects the presence beneath the scanner of the object carrying the characters. Initially, the digital signal representing the black/white state of successive points along each scan line is passed to a logic circuit, Fig. 2 (not shown), comprising counters and gates. The circuit counts successive black points and, if the number of these is sufficient to indicate that mark 16 is being scanned, a second counter counts successive white points. If the number of white points counted approximates the known distance D between marks 16 and 17 black marks are again counted to detect mark 17. If three successive scans are found to have traversed both marks the subsequent signals from the scanner are gated to a processor. The processor can determine the skew angle α by noting the vertical height S of any part of mark 17 lying above the top of mark 16, or vice versa in the case of opposite skew. The processor identifies each character in turn. A character to be recognized may be considered as lying on an imaginary rectangular grid, Fig. 7, each horizontal row of grid elements corresponding to one scan line. As disclosed in Specification 1,359,781 eight groups of horizontal and vertical sections of the character are employed in the recognition process. Each vertical group 1-3 and horizontal group 4-8 consists of two or more sections (rows or columns), e.g., 1 1 , 1 2 , 1 3 , and in each group the section having the most black points is selected to provide a single bit, the resulting 8-bit signal representing the character despite any small amount of skew. When the skew is larger than can be accommodated by the technique disclosed in Specification 1,359,781 the grid is in effect distorted by the processor. Thus in the case of the skewed character shown in Fig. 7 instead of considering the vertical columns 1 1 , 1 2 , 1 3 , three stepped columns as indicated by 54, 55 are considered in order to determine the state of the bit corresponding to group 1. Each other group consists of similarly stepped sections. The size of the steps depends on the skew angle α.