Method and device for recognition of delivery data on mail matter

• US 6,577,749 B1
• Filed: 03/27/2000
• Issued: 06/10/2003
• Est. Priority Date: 09/27/1997
• Status: Expired due to Fees

• Current Assignee: Siemens AG
• Original Assignee: Siemens AG
• Inventors: Rosenbaum, Walter
• Primary Examiner(s): Johns, Andrew W.
• Assistant Examiner(s): NAKHJAVAN, SHERVIN K
• Application Number:
  US09/509,385
• Time in Patent Office: 1,170 Days
• Field of Search: 382/101, 382/102, 382/182, 382/229, 382/231, 382/309, 382/310, 382/311, 382/317, 382/321, 379/910, 345/716, 345/727, 345/728, 345/978, 704/239, 704/236, 704/237, 704/251, 704/252
• US Class Current: 382/101
• CPC Class Codes: B07C 3/20 Arrangements for facilitating the visual reading of addresses, e.g. display arrangements coding stations
• CPC Class Codes: G06V 10/987 with the intervention of an operator

Abstract

The invention relates to a method and a device for detecting delivery data on mail items with the aid of video coding stations (40), to which the recorded, digitized and stored images of the mail item surfaces are transmitted, following an ambiguous automatic evaluation of the delivery data in an OCR unit (100). In the process, at least the names are entered by means of a voice-input unit (60, 70). The voice recognition candidates are then selected with the aid of the ambiguous results from the automatic OCR evaluation (120). It is advantageous if this occurs through a comparison of the respectively determined number of characters or by means of a position-related comparison of the characters for the candidates with the characters determined in the OCR unit and by taking into account the reliabilities.

Claims

1. A method for recognizing delivery data on mail items with the aid of video coding stations, to which the recorded, digitized and stored images of the mail item surfaces containing the delivery data have been transmitted following an ambiguous, automatic evaluation of the delivery data in at least one OCR unit, including a comparison with a directory, said method involving the input of at least the names of the respective delivery data, displayed on the respective video coding station monitor, by means of a voice input unit that comprises a microphone and a voice recognition module, wherein the voice recognition candidates are selected with the incomplete and/or ambiguous results of the automatic OCR evaluation.

2. A method according to claim 1, wherein the numerical components of the displayed delivery data are input with the aid of a keyboard.

3. A method according to claim 1, wherein the number of characters is determined in the OCR evaluation as accurately as possible and is compared to the number of corresponding characters for the voice recognition candidates and wherein those candidates are sorted out, for which the number of characters deviates by more than one statistically determined limit value from the number of characters detected in the OCR evaluation.

4. A method according to claim 3, wherein the number of characters is determined from the result of the segmenting of the OCR evaluation.

5. A method according to claim 1, wherein the characters determined in the OCR evaluation, to which reliabilities have been assigned, are compared position-related to the characters for the voice recognition candidates and wherein the candidate with the most reliable match above a statistically determined limit value is selected.

6. A method according to claim 1, wherein for the determination of a street name,

first, the street name candidates with the associated house numbers that have been input are selected on the basis of their existence in a street directory,

second, the remaining candidates are then selected based on the number of characters as compared to the determined number of characters from the OCR evaluation and

third, the remaining candidates are selected in a position-related comparison of the characters for the candidates with the characters from the OCR evaluation.

7. A method according to claim 1, wherein a search is conducted with the aid of a number OCR unit in all possible ROI address fields for the numerical components of the delivery data that was input via keyboard and wherein the resulting numbers are used to determine the searched-for address line and its orientation.

8. A method according to claim 7, wherein the name field in the address line is determined with the aid of the numbers found with the number OCR unit and the position of said numbers.

9. A device for detecting delivery data on mail items, comprising

a device for obtaining images of mail items (17),

an OCR unit (20, 100) for the automatic evaluation of the images of mail item surfaces containing the delivery data, including a comparison with a directory,

a device for video coding the images of the mail item surfaces containing the delivery data, said device having at least one video coding station (40) with respectively one keyboard (50), one monitor and one voice input unit that consists of a microphone (60) and a voice recognition module (70),

a processing unit (30) for controlling the data flows between the input units and the output units of the device for video coding and the OCR unit (20, 100), wherein the candidates from the respective voice recognition module (70) are selected in the processing unit (30) with the aid of the incomplete and/or ambiguous results of the OCR unit (20, 100).

10. A device according to claim 9, wherein the number of characters determined in the OCR unit (20, 100) is compared in the processing unit (30) to the number of characters for the candidates from the respective voice recognition module (70) and wherein those candidates are sorted out, for which the character number deviates by more than one statistically determined limit value from the number of characters of the OCR unit (20, 100).

11. A device according to claim 9, wherein the characters determined in the OCR unit (20, 100), which have been provided with reliabilities, are compared in the processing unit (30) position-related to the characters for the candidates from the respective voice recognition module (70) and wherein the candidate that coincides with the highest reliability above a statistically determined limit value is selected.