 Method for enabling privacy and trust in electronic communities |
| A major impediment to using recommendation systems and collective knowledge for electronic commerce ... |
|
| Method and apparatus for providing secure access to a computer system resource |
| One illustrative embodiment of the invention is directed to a method of managing access to one of a ... |
|
| Method for estimating the price per square foot value of real property |
| OF THE INVENTION This invention estimates the dollar per square foot value of a subject property ... |
|
| Net-value creation and allocation in an electronic trading system |
| The present invention describes a system, method, and apparatus for Value Creation in an electronic ... |
|
| Data recording system |
| It is an object of the invention so to design a system of the kind described wherein, after the ... |
|
| Automated apparatus and method for object recognition at checkout counters |
| What is claimed is as follows: 1. A method for automatically recognizing and identifying, without ... |
|
| System for dispensing dry agricultural chemicals |
| FIG. 1 shows a simplified diagram of a planter 20 incorporating a distributed control material ... |
|
| Licensee notification system |
| An object of the present invention is to provide a licensee notification system whose psychological ... |
|
| System and method for product registration |
| In accordance with the present invention, a system and method for product registration and post-... |
|
| Method and system for delivering digital products electronically |
| A system and method for delivering a digital product electronically is disclosed. The system and ... |
|
|
Method for exploiting correlated mail streams using optical character recognition
| Details |
Inventors: Gozzo, Francesco; Higgins, Joseph T.;
Assignee: Lockheed Martin Corporation (Bethesda, MD)
Primary Examiner: Couso; Jose L.
Assistant Examiner: Do; Anh Hong
Attorney, Agent or Firm: McGuireWoods, LLP
A method for exploiting correlated mail streams using optical character recognition is provided in which a static database is used to store data based on training. Real-time data for the parameters of interest, such as address block location, zip code, city, state, and font size or type is collected from the mail processing equipment in order to generate a statistical information database. The dynamic database can include probability density functions, correlations statistics, mean, variance, and high order moments. The statistical parameters are tracked using recursive least squares schemes with various windowing options, as will as moving average linear filters. Based on cost models which indicate the cost of making various types of errors in the OCR process, a decision threshold is determined which is based on the real-time statistics of the mail stream. The decision threshold determines the confidence level required by the adaptive process in order to assign a previously rejected mail pieces based solely on correlation statistics. The decisions threshold will adapt to the statistics of the mail stream and is not a constant value. Previously unassigned characters are assigned according to the decision threshold determination and assignment processes. |
|
DETAILED DESCRIPTION OF THE INVENTION Before describing the improvements made by use of adaptive OCR, a Remote Computer Reader (RCR) prototype system which is used to process envelope images which are rejected by the current state-of-the-art postal technology will first be described.
Referring now to the drawing where like numerals indicate like elements throughout and more particularly FIG.
1 thereof, a flow diagram of the steps involved in processing an image using an RCR system is shown.
After an initial decompression stage at step 31, the image is processed sequentially by an address block location function at step 33, character segmentation function at step 35, recognition and address interpretation functions at steps 37 and 41, respectively.
Since the incoming image to step 31 is compressed, a decompression process must be performed on the image.
The compression algorithm is based on the CCITT recommendation for the transmission of documents by facsimile.
This algorithm uses a modified Huffman code to one-dimensionally compress every fourth (k=4) line.
The other lines are two-dimensionally compressed based on the previous one-dimensionally compressed line.
The next step 33 in the process is to find the destination Address Block Location (ABL).
ABL is a complex and time consuming step because there are so many different things that appear on envelopes.
For example, the return address block, stamp and advertising logos can all look like a possible ABL candidate.
Referring now to FIG.
2, the Address Block Location function in step 33 of FIG.
1 is shown in more detail.
The input to ABL is the full resolution image from which a low resolution image is created at step 43.
The low-resolution image is used to determine possible destination ABLs in step 45.
These candidate address blocks (CABs) are then classified as either hand, indicating hand written characters, or machine, indicating machine created characters and then ranked in step 47.
The top three hand and machine CABs are then passed on to the segmentation process, shown in step 35 in FIG
|
|
Finance 
