Boundary trace slope feature detection system

• US 4,326,190 A
• Filed: 08/30/1978
• Issued: 04/20/1982
• Est. Priority Date: 08/30/1978
• Status: Expired due to Term

• Current Assignee: Bantec Incorporated
• Original Assignee: David L. Borland, Lyle W. Shaw
• Inventors: Shaw, Lyle W., Borland, David L.
• Primary Examiner(s): Boudreau, Leo H.
• Application Number:
  US05/938,271
• Time in Patent Office: 1,329 Days
• Field of Search: 340/146.3 AE, 340/146.3 AC, 340/146.3 MA, 364/416
• US Class Current: 382/197
• CPC Class Codes: G06V 10/20 Image preprocessing
• CPC Class Codes: G06V 10/46 Descriptors for shape, contour or point-related descriptors, e.g. scale invariant feature transform [SIFT] or bags of words [BoW]; Salient regional features colour feature extraction G06V10/56

Abstract

A character feature detection system is provided for reliably reading alphanumeric characters. An information field is scanned by sensor array and the sensor responses are digitized to form a binary character image. The image is traced from boundary point to boundary point. A transition from one boundary point to another is defined by one of eight equally divergent vectors which are indicative of the local slope of a boundary between connected points. The sequence of vectors in their xy locations recorded in the boundary slopes are indicated by averaging vectors representative of the cell to cell transition. Character features are detected from the sequence of boundary slopes to form a feature set. The feature decision logic analyzes the primary features to form a set of secondary features based upon the occurrence and order of the features. The set of secondary features are then analyzed in a sequential logic decision tree to identify the character.

Claims

1. An improved method of recognizing descriptive features of characters, which have been scanned with a plurality of sensor cells, from binary representation thereof stored in a memory means, and tracing around the stored character representations, comprising the steps of

(a) deriving initial features of a character such as circumference count prior to tracing around the character;

(b) generating a single boundary trace on the boundary of the character to locate character features;

(c) defining the contours of the character without regard to size of the character based upon the character shape and features derived during the boundary trace;

(d) Weighting the derived features; and

(e) generating a character code based upon the weighted features of the character.

2. The method according to claim 1 including the step of generating during the single boundary trace first signals representing the outer boundaries and inner loops of said character by a vector slope analysis means in communication with said memory means.

3. The method according to claim 2 including the steps of generating second signals which are vector representations of said boundaries by a trace control means receiving said first signals, each of said representations comprising one of a set of a plurality of equal divergent vectors and generating a set of indicia representative of said characters based upon averages of selected sets of said second signals by said trace control means.

4. An improved method of recognizing descriptive features of characters derived from tracing around stored character representation, which have been scanned with a plurality of sensor cells, from binary representations thereof stored in a memory means comprising the steps of:

(a) deriving the initial features of circumference, area and stroke width of each character.

(b) generating without regard to size during a single trace weighted features representative to the character;

(c) determining the slope of the generated features; and

(d) deriving a character code based upon the weighted features, slope of the features and sequence of the features representative of the character.

5. An improved method of recognizing descriptive features of characters by tracing around the circumference thereof, which had been scanned with a plurality of sensor cells, from binary representations thereof stored in a memory means comprising the steps of:

(a) generating an initial set of weighted features representative of the scanned character including stroke width, area and measurement of the circumference prior to tracing around the character;

(b) tracing around the character a single time to detect a set of roadmap features;

(c) calculating feature data based upon the initial features and the roadmap features;

(d) storing the calculated feature data; and

(e) presenting the data to a decision tree to make a character decision.