US5712922A
Expired / Lapsed
US
⏱️ 0.0 years remaining
Neural network optical character recognition system and method for classifying characters in a moving web
Assignee: Eastman Kodak Co
•
Filing: 1994-11-15
•
Granted: 1998-01-27
•
Expires: 2014-11-15
Abstract
Abstract A neural network based optical character recognition technique is presented for identifying characters in a moving web. Image acquisition means defines an imaging window through which the moving web passes such that the characters printed thereon can be imaged. Classification data is extracted and accumulated for each printed web character passing through the imaging window. A light source provides transmissive illumination of the web as it is being imaged. A neural network accelerator is coupled to the image acquisition means for intelligent processing of the accumulated classification data to produce therefrom printed character classification information indicative of each corresponding character imaged. A processor is coupled to the accelerator for converting the classification information into the appropriate ASCII character code. The technique is particularly useful for reading dot-matrix-type characters on a noisy, semi-transparent background at fast real-time rates. A neural network algorithm based recognition method is also described.
Inventors
- 👤 Kenneth H. Loewenthal
- 👤 Steven M. Bryant
Bibliographic Details
Application Number
—
Jurisdiction
US
Data Source
google_patents
Last Synchronized
2026-11-09 14:38
CPC Classifications
G; G06; G06V; G06V30/00; G06V30/10; G06V30/14; G06V30/1434; G06V30/19; G06V30/192; G06V30/194
Claims Analysis
Total Claims: 39
Claim 1
Independent Claim
276 words
A neural network based optical character recognition system for extracting and classifying characters disposed on a moving web of material, said recognition system comprising: image acquisition means defining an imaging window, said image acquisition means being positionable such that said characters pass through said defined imaging window, said image acquisition means extracting and accumulating image data on a web character passing through said imaging window; a light source for providing illumination of said web character as said web character passes through said defined imaging window; neural network accelerator means coupled to said image acquisition means for processing accumulated image data on said web character passing through said imaging window according to a predefined multilayer neural network, and for producing therefrom character classification information, said neural network accelerator means including: dual-ported memory means receiving said accumulated image data from said image acquisition means, said dual-ported memory means having a first portion and a second portion, said first portion of said dual-ported memory means initially storing said accumulated image data, and n processing nodes connected in parallel to receive from said first portion and simultaneously process image data at different nodes for an input layer in said multilayer neural network, an output of each processing node for said input layer being stored in said second portion of said memory means, n being an integer≧1, said first portion and second portion subsequently storing processing node outputs for alternate layers of said multilayer neural network, respectively; and classification means coupled to said neural network accelerator means for classifying said web character imaged by said image acquisition means using the corresponding character classification information produced by said neural network accelerator means.
Claim 2
Dependent Claim
276 words
A neural network based optical character recognition system for extracting and classifying characters disposed on a moving web of material, said recognition system comprising: image acquisition means defining an imaging window, said image acquisition means being positionable such that said characters pass through said defined imaging window, said image acquisition means extracting and accumulating image data on a web character passing through said imaging window; a light source for providing illumination of said web character as said web character passes through said defined imaging window; neural network accelerator means coupled to said image acquisition means for processing accumulated image data on said web character passing through said imaging window according to a predefined multilayer neural network, and for producing therefrom character classification information, said neural network accelerator means including: dual-ported memory means receiving said accumulated image data from said image acquisition means, said dual-ported memory means having a first portion and a second portion, said first portion of said dual-ported memory means initially storing said accumulated image data, and n processing nodes connected in parallel to receive from said first portion and simultaneously process image data at different nodes for an input layer in said multilayer neural network, an output of each processing node for said input layer being stored in said second portion of said memory means, n being an integer≧1, said first portion and second portion subsequently storing processing node outputs for alternate layers of said multilayer neural network, respectively; and classification means coupled to said neural network accelerator means for classifying said web character imaged by said image acquisition means using the corresponding character classification information produced by said neural network accelerator means.
Claim 3
Dependent Claim
47 words
The recognition system of claim 1, wherein said image acquisition means comprises a line scan CCD camera with a view lens which serves to define said imaging window, and wherein said web passes through said defined imaging window substantially orthogonal to an optical axis of said camera.
Claim 4
Dependent Claim
19 words
The recognition system of claim 2, wherein said multilayer neural network comprises a fully interconnected, multilayer feedforward neural network.
Claim 5
Dependent Claim
51 words
The recognition system of claim 1, further comprising means for switching inputs to said n processing nodes from said first portion of said dual-ported memory means to said second portion thereof once processing of accumulated image data of all nodes in said input layer of said multilayer neural network is completed.
Claim 6
Dependent Claim
39 words
The recognition system of claim 4, wherein the output of each node during processing of data by the n nodes of a second layer of said multilayer neural network is overwritten into said first portion of said memory means.
Claim 7
Dependent Claim
90 words
The recognition system of claim 5, wherein the outputs of said parallel connected processing nodes for each layer of said multilayer neural network are stored in a selected one of said first portion and said second portion of said dual-ported memory means, the selected memory means portion being other than said portion of said dual-ported memory means containing the input data to the parallel processing nodes for said layer, said input data for said layer comprising the stored processing node outputs for the previous layer in the multilayer neural network.
Claim 8
Dependent Claim
31 words
The recognition system of claim 1, wherein the outputs of said processing nodes are passed through sigmoid function processing means for data squashing prior to being stored in said memory means.
Claim 9
Dependent Claim
34 words
The recognition system of claim 1, further comprising a dedicated, parallel-configured bus for transferring said accumulated image data from said image acquisition means to said dual-ported memory means of said neural network accelerator means.
Claim 10
Dependent Claim
55 words
The recognition system of claim 8, further comprising a multiplexer and a system data bus, said multiplexer being disposed within said neural network accelerator means and connected to each of said dedicated parallel input bus and said system data bus for multiplexing of image data from either one of said buses into said memory means.
Claim 11
Dependent Claim
71 words
The recognition system of claim 1, wherein each of said n parallel connected processing nodes comprises a multiplier/accumulator logic circuit having two inputs and one output, a first one of said inputs being connected to said memory means and a second one of said inputs being connected to a unique system memory address which contains predefined weighting factors for corresponding processing nodes in the respective layers of the multilayer neural network.
Claim 12
Dependent Claim
33 words
The recognition system of claim 1, further comprising control logic for controlling the flow of data within said neural network accelerator means, said neural network accelerator means comprising a neural network accelerator board.
Claim 13
Dependent Claim
47 words
The recognition system of claim 1, further comprising means for preprocessing said image data accumulated by said image acquisition means prior to processing of said data by said neural network accelerator means, said preprocessor means including means for noise filtering and means for formatting said image data.
Claim 14
Dependent Claim
63 words
The recognition system of claim 12, wherein said noise filtering means and formatting means of said preprocessor means include: thresholding means for filtering invalid image data information; dilation and compression means for simultaneously representing said accumulated image data in a dilated compressed form; and means for converting said compressed image data into a predefined format for acceptance by said neural network accelerator means.
Claim 15
Dependent Claim
40 words
The recognition system of claim 13, wherein said simultaneous dilation and compression means includes a plurality of multiple input OR gates, each of said OR gates being connected to receive input pixels from scan lines of said image acquisition means.
Claim 16
Dependent Claim
36 words
The recognition system of claim 14, wherein each of said OR gates is connected to receive four input pixels from adjacent scan lines, two of said input pixels being obtained from each of said scan lines.
Claim 17
Dependent Claim
47 words
The recognition system of claim 12, wherein said image acquisition means includes a frame grabber memory means, said frame grabber memory means being coupled to an input of said preprocessor means, said frame grabber memory means receiving accumulated image data to be preprocessed by said preprocessing means.
Claim 18
Dependent Claim
40 words
The recognition system of claim 1, wherein said moving web of material comprises a moving web of photographic material and wherein said light source provides infrared transmissive illumination to said web as said web passes through said defined imaging window.
Claim 19
Dependent Claim
112 words
The recognition system of claim 1, wherein said characters disposed on said moving web of material comprise edgemarkings, each of said edgemarkings being assembled into one of a plurality of groupings each of which is indexed with a special message character, the distance in each grouping to each character from said respective special message character being known, and wherein said system further comprises: means for identifying said special message characters; and encoder means for clocking the web distance from each identified special message character, said encoder means providing a collect image trigger signal to said image acquisition means once a known distance to a character from said special message character is reached.
Claim 20
Dependent Claim
24 words
The recognition system of claim 18, wherein said encoder means includes an optical shaft encoder disposed in communication with said moving web of material.
Claim 21
Dependent Claim
59 words
The recognition system of claim 1, wherein said classification means includes a processor for formatting said classification information produced by said neural network accelerator means into ASCII code, and wherein said system further comprises communication means for facilitating transfer of said ASCII coded classification information to a host computer for at least one of process accounting and process control.
Claim 22
Dependent Claim
35 words
The recognition system of claim 1, wherein said system further comprises at least one of display means for viewing said classified web characters, and storage means for storing said classification information on said web characters.
Claim 23
Dependent Claim
14 words
The recognition system of claim 1, wherein said characters comprise dot matrix-type printed characters.
Claim 24
Dependent Claim
39 words
The recognition system of claim 22, wherein said printed characters to be extracted and classified are edgemarked on said web and comprise at least one of machine information, roll information and footage information for said moving web of material.
Claim 25
Dependent Claim
30 words
The recognition system of claim 23, wherein said web of material comprises photographic material, and wherein said dot matrix print characters are semi-transparent and wherein said photographic material is semi-transparent.
Claim 26
Dependent Claim
14 words
The recognition system of claim 24, wherein said characters comprise machine printed numeric characters.
Claim 27
Independent Claim
275 words
A method for extracting and classifying characters disposed on a moving web of material using a multilayer neural network, said method comprising the steps of: (a) passing said moving web through a predefined imaging window; (b) illuminating said web within said imaging window simultaneous with said passing step (a); (c) acquiring an image of said web within said imaging window simultaneous with said passing step (a); (d) extracting and accumulating image data on characters disposed on said web imaged in said step (c); and (e) utilizing a multilayer neural network to process image data accumulated in said step (d), said multilayer neural network producing character classification information based upon accumulated image data, said utilizing step including: for each layer of said multilayer neural network, processing data by n nodes thereof in parallel, n being an integer≦1; utilizing a memory storage having a first portion and a second portion for simultaneously providing input data and storing output data for each respective layer of said multilayer neural network; initially storing in the first portion of said memory storage said image data accumulated in said step (d); during processing of accumulated image data received from said first portion by n nodes of an input layer of said multilayer neural network, storing the output of said n nodes into said second portion of said memory storage; and for each network layer subsequent said input layer, alternately coupling one of said second portion and said first portion of the memory storage as an input to said n nodes, and storing the output of said n nodes into the other of said memory storage portions not containing the input data thereto.
Claim 28
Independent Claim
275 words
A method for extracting and classifying characters disposed on a moving web of material using a multilayer neural network, said method comprising the steps of: (a) passing said moving web through a predefined imaging window; (b) illuminating said web within said imaging window simultaneous with said passing step (a); (c) acquiring an image of said web within said imaging window simultaneous with said passing step (a); (d) extracting and accumulating image data on characters disposed on said web imaged in said step (c); and (e) utilizing a multilayer neural network to process image data accumulated in said step (d), said multilayer neural network producing character classification information based upon accumulated image data, said utilizing step including: for each layer of said multilayer neural network, processing data by n nodes thereof in parallel, n being an integer≦1; utilizing a memory storage having a first portion and a second portion for simultaneously providing input data and storing output data for each respective layer of said multilayer neural network; initially storing in the first portion of said memory storage said image data accumulated in said step (d); during processing of accumulated image data received from said first portion by n nodes of an input layer of said multilayer neural network, storing the output of said n nodes into said second portion of said memory storage; and for each network layer subsequent said input layer, alternately coupling one of said second portion and said first portion of the memory storage as an input to said n nodes, and storing the output of said n nodes into the other of said memory storage portions not containing the input data thereto.
Claim 29
Dependent Claim
20 words
The method of claim 26, further comprising the step of processing said character classification information into a corresponding ASCII code.
Claim 30
Dependent Claim
25 words
The method of claim 26, further comprising the step of predefining said neural network utilized in said step (e) using a multilayer back-propagation training technique.
Claim 31
Dependent Claim
27 words
The method of claim 26, wherein said step (c) comprises acquiring an optical image of said web within said imaging window simultaneous with said passing step (a).
Claim 32
Dependent Claim
16 words
The method of claim 26, wherein said neural network comprises a fully interconnected, multilayer neural network.
Claim 33
Dependent Claim
35 words
The method of claim 26, further comprising the step of preprocessing said image data prior to said step (e), said preprocessing including substantially simultaneously dilating and compressing said image data to a dilated compressed form.
Claim 34
Dependent Claim
24 words
The method of claim 31, wherein said simultaneous dilation and compression includes passing said image data through a plurality of parallel disposed OR gates.
Claim 35
Dependent Claim
30 words
The method of claim 26, wherein said web comprises a photographic material and said illuminating step (b) includes illuminating said web within said predefined imaging window with infrared transmissive lighting.
Claim 36
Dependent Claim
182 words
A neural network accelerator for processing input data in a feed forward operational mode according to a predefined multilayer neural network, said accelerator comprising: memory means divided into a first memory portion and a second memory portion, said input data being initially received and stored by said first memory portion of said memory means; n parallel connected processing nodes for receiving and processing as a first network layer said input data stored in said first memory portion, each of said processing nodes being coupled to provide first layer output data to said second memory portion of said memory means for storage, n being an integer≦1; and means for switching processing node inputs from said first memory portion of said memory means to said second memory portion of said memory means upon completion of said input data processing as said first network layer by said n parallel connected processing nodes, whereby the first layer output data stored by said second memory portion is input from said second memory portion to said n parallel connected processing nodes for processing as a second network layer.
Claim 37
Dependent Claim
182 words
A neural network accelerator for processing input data in a feed forward operational mode according to a predefined multilayer neural network, said accelerator comprising: memory means divided into a first memory portion and a second memory portion, said input data being initially received and stored by said first memory portion of said memory means; n parallel connected processing nodes for receiving and processing as a first network layer said input data stored in said first memory portion, each of said processing nodes being coupled to provide first layer output data to said second memory portion of said memory means for storage, n being an integer≦1; and means for switching processing node inputs from said first memory portion of said memory means to said second memory portion of said memory means upon completion of said input data processing as said first network layer by said n parallel connected processing nodes, whereby the first layer output data stored by said second memory portion is input from said second memory portion to said n parallel connected processing nodes for processing as a second network layer.
Claim 38
Dependent Claim
40 words
The neural network accelerator of claim 34, wherein the outputs of said n parallel connected processing nodes during data processing as a second layer of said multilayer neural network are overwritten into said first memory portion of said memory means.
Claim 39
Dependent Claim
49 words
The neural network accelerator of claim 35, wherein said switching means further comprises means for switching the processing node inputs from said second memory portion to said first memory portion for data processing by said n parallel connected processing nodes as a third layer of said multilayer neural network.
References Cited (Backward Citations)
US4414685A
Cited
US4680805A
Cited
US5068799A
Cited
EP0288322B1
Cited
USH681H
Cited
US5014327A
Cited
US4994982A
Cited
US5506915A
Cited
US5329630A
Cited
US5434951A
Cited
US4932065A
Cited
US5048100A
Cited
US4974169A
Cited
US5553196A
Cited
US5267185A
Cited
US5060278A
Cited
US5426745A
Cited
US5105387A
Cited
JPH03152406A
Cited
US5067164A
Cited
US5504838A
Cited
US5165009A
Cited
US5048097A
Cited
US5040229A
Cited
US5054094A
Cited
US5052043A
Cited
US5398300A
Cited
US5194958A
Cited
US5204914A
Cited
US5465375A
Cited
USPTO Annuity & Maintenance Windows
| Annuity Stage | Window Opens | Payment Deadline | Surcharge Period Until | Estimated Fee | Status |
|---|---|---|---|---|---|
| 3.5 Years (Year 4) | 2001-01-26 | 2001-07-28 | 2002-01-27 | $2000.00 | Window Closed / Lapsed |
| 7.5 Years (Year 8) | 2005-01-26 | 2005-07-28 | 2006-01-27 | $3760.00 | Window Closed / Lapsed |
| 11.5 Years (Year 12) | 2009-01-26 | 2009-07-28 | 2010-01-27 | $7700.00 | Window Closed / Lapsed |