Pattern recognition system
First Claim
1. A pattern recognition system for recognizing plural determined two-dimensional patterns located at different unknown positions within a larger two dimensional image area, said system comprising:
- means for the input of electronic data representing a two-dimensional area of digitized pixel areas, each of said pixel areas having a plurality of contiguous pixels corresponding to a portion of said image area which remains unsegmented with respect to the positions of the plural patterns included therewithin to be recognized;
analysis means for analyzing successive potential character-occupying portions of said data associated with each of successive pixel areas of said unsegmented image area and for assigning to each of said pixel areas, in dependence on the corresponding said portion of data, a corresponding one label of a reference set of labels representing one of said predetermined patterns and a corresponding confidence measure indicating a degree of confidence in the assignment of said corresponding one label;
buffer means for storing area values, said area value corresponding to a plurality of said successive overlapping pixel areas forming a two dimensional array of pixel areas, said area values comprising, for each of said neighboring pixel areas, positional information defining the position of the associated pixel area in said two dimensional image and said assigned label and a corresponding confidence measure assigned to the area by said analysis means, anddecision means for outputting as electronic signals, labels and positional information corresponding to those pixel areas in said two dimensional array whose corresponding confidence measures indicate a confidence in the correctness of assignment of their corresponding labels which is in a predetermined relationship with that indicated by each of the confidence measures of other members of a two-dimensional overlapping sub-array of neighboring pixel areas within said two-dimensional array, the said relationship being such that the said labels and positional information correspond to those pixel areas whose confidence measures are maxima within each said two-dimensional sub-array.
1 Assignment
0 Petitions
Accused Products
Abstract
A pattern recognition system comprising pattern scanning data input means (1), a dichotomizing device (3), a feature matching device (5), a label assignment circuit (7) and a syntactic decision processing device (9). The label assignment circuit assigns labels and confidence measures to data corresponding to successive areas of the pattern. The syntactic decision processing device (9) comprises a buffer for storing area values comprising positional information, corresponding labels and confidence measures for adjacent areas in the pattern forming a two dimensional array and also comprises decision means for outputting labels and positional information for those areas whose confidence measures form two dimensional maxima with respect to those for neighboring areas in the array. The system may operate continuously as area values corresponding to successive portions of input data are added to the buffer.
-
Citations
18 Claims
-
1. A pattern recognition system for recognizing plural determined two-dimensional patterns located at different unknown positions within a larger two dimensional image area, said system comprising:
-
means for the input of electronic data representing a two-dimensional area of digitized pixel areas, each of said pixel areas having a plurality of contiguous pixels corresponding to a portion of said image area which remains unsegmented with respect to the positions of the plural patterns included therewithin to be recognized; analysis means for analyzing successive potential character-occupying portions of said data associated with each of successive pixel areas of said unsegmented image area and for assigning to each of said pixel areas, in dependence on the corresponding said portion of data, a corresponding one label of a reference set of labels representing one of said predetermined patterns and a corresponding confidence measure indicating a degree of confidence in the assignment of said corresponding one label; buffer means for storing area values, said area value corresponding to a plurality of said successive overlapping pixel areas forming a two dimensional array of pixel areas, said area values comprising, for each of said neighboring pixel areas, positional information defining the position of the associated pixel area in said two dimensional image and said assigned label and a corresponding confidence measure assigned to the area by said analysis means, and decision means for outputting as electronic signals, labels and positional information corresponding to those pixel areas in said two dimensional array whose corresponding confidence measures indicate a confidence in the correctness of assignment of their corresponding labels which is in a predetermined relationship with that indicated by each of the confidence measures of other members of a two-dimensional overlapping sub-array of neighboring pixel areas within said two-dimensional array, the said relationship being such that the said labels and positional information correspond to those pixel areas whose confidence measures are maxima within each said two-dimensional sub-array. - View Dependent Claims (2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16)
-
-
17. A pattern recognition system in which the relative positions of plural patterns to be recognized within input data are automatically determined as a part of a pattern recognition analysis procedure thereby avoiding the need to preprocess and/or segment the incoming data to initially determine such positions prior to pattern recognition analysis, said system comprising:
-
data input means for supplying a continuous stream of digitized binary-valued data representing successive data values of a raster-scanned array of data values containing plural of predetermined patterns therein and with said data values defining a two-dimensional grid of x,y coordinates overlaying said array; feature matching means connected to receive said continuous data and to continuously identify the existence of any of N predetermined features located within an area of P×
Q data values located at successive X,Y coordinates within said array and to produce an N-dimensional feature vector for each such X,Y coordinate;label assignment means connected to continuously receive said N-dimensional feature vectors and to continuously assign one of a predetermined set of M pattern labels to each such feature vector together with an accompanying confidence factor for each such X,Y coordinate; and syntactic processing means connected to continuously receive said assigned pattern labels and accompanying confidence factors for each such X,Y coordinate and, by comparing such assigned labels and confidence factors for one X,Y coordinates so as to locate local maxima of confidence factors therein, to output recognized pattern data representing the sequence of predetermined patterns recognized in said array together with their relative position therewithin.
-
-
18. A pattern recognition method in which the relative positions of plural patterns to be recognized within input data are automatically determined as a part of a pattern recognition analysis procedure thereby avoiding the need to preprocess and/or segment the incoming data to initially determine such positions prior to pattern recognition analysis, said method comprising the steps of:
-
supplying a continuous stream of digitized binary-valued data representing successive data values of a raster-scanned array of data values containing plural of predetermined patterns therein with said data values defining a two-dimensional grid of x,y coordinated overlaying said array; continuously identifying the existence of any of N predetermined features located within an area of P×
Q data values located at successive X,Y coordinates within said array and producing an N-dimensional feature vector for each such X,Y coordinate;continuously assigning one of a predetermined set of M pattern labels to each such feature vector together with an accompanying confidence factor for each such X,Y coordinate; and continuously receiving said assigned pattern labels and accompanying confidence factors for each such X,Y coordinate and, by comparing such assigned labels and confidence factors for one X,Y coordinate to those associated with neighboring X,Y coordinates so as to locate local maxima of confidence factors, outputting recognized pattern data representing the sequence of predetermined patterns recognized in said array together with their relative positions therewithin.
-
Specification