US 12,254,682 B2
Dual stage neural network pipeline systems and methods
Xi Chen, Annandale, VA (US); Mamadou Diallo, San Francisco, CA (US); and Qiang Xue, Vienna, VA (US)
Assigned to CAPITAL ONE SERVICES, LLC, McLean, VA (US)
Filed by Capital One Services, LLC, McLean, VA (US)
Filed on Aug. 5, 2021, as Appl. No. 17/394,594.
Application 17/394,594 is a continuation of application No. 16/705,604, filed on Dec. 6, 2019, granted, now 11,126,892.
Application 16/705,604 is a continuation of application No. 16/281,971, filed on Feb. 21, 2019, granted, now 10,558,894, issued on Feb. 11, 2020.
Application 16/281,971 is a continuation of application No. 15/904,964, filed on Feb. 26, 2018, granted, now 10,262,235, issued on Apr. 16, 2019.
Prior Publication US 2021/0365737 A1, Nov. 25, 2021
This patent is subject to a terminal disclaimer.
Int. Cl. G06K 9/00 (2022.01); G06F 18/214 (2023.01); G06N 3/045 (2023.01); G06N 3/08 (2023.01); G06V 10/26 (2022.01); G06V 10/82 (2022.01); G06V 20/62 (2022.01); G06V 30/148 (2022.01); G06V 30/19 (2022.01); G06V 30/242 (2022.01); G06V 30/32 (2022.01)
CPC G06V 10/82 (2022.01) [G06F 18/2148 (2023.01); G06N 3/045 (2023.01); G06N 3/08 (2013.01); G06V 10/273 (2022.01); G06V 20/62 (2022.01); G06V 20/63 (2022.01); G06V 30/153 (2022.01); G06V 30/19173 (2022.01); G06V 30/242 (2022.01); G06V 30/36 (2022.01); G06V 20/625 (2022.01)] 19 Claims
OG exemplary drawing
 
1. A method of identifying and recognizing characters within a magnetic ink character recognition (MICR) line, the method comprising:
receiving an image comprising image data; and
recognizing, using a neural network (NN), an MICR line within the image,
wherein recognizing the MICR line comprises generating an NN resultant matrix, using the NN and based on the image data and recognizing the MICR line within the NN resultant matrix by:
summing up one or more row portions of the NN resultant matrix to determine a plurality of detection counts, wherein each one of the one or more row portions is associated with a respective detection count; and
comparing each of the plurality of detection counts to a predetermined threshold to determine which of the one or more row portions contain data indicative of an MICR line, and wherein the predetermined threshold is determined based in part on a number of consecutive rows within the one or more row portions having a nonzero detection count.