License plate recognition
AssigneeQ-Free Netherlands B.V.
InventorJos A. G. Nijhuis
Life of this patent
Estimating the term…
Abstract
A license plate recognition and secondary image review system and processes are described. The system supplements optical character recognition with a secondary image review process that matches an acquired image with either images in a trusted database of historic confirmed identification or with fingerprint data derived from the images compared with confirmed fingerprint data from prior identifications contained in a trusted data base. The system provides an alternative to manual reviews in a selected subset of OCR identifications, while maintaining a preselected accuracy and risk of misidentification.
Claims4
A computer-implemented method for identifying character strings on a target license plate said method comprising:
- a) a learning process comprising:
- i) acquiring a plurality of images of a plurality of vehicles on a roadway, said plurality of images each including a license plate, the license plates including the target license plate and non-target license plates,
- ii) identifying, solely by optical character recognition, a character string on the license plates in each of the plurality of images, and, grouping the images identified as having a same character string solely by optical character recognition, thereby sorting the images into a plurality of first identification groups each having an associated character string and a plurality of images,
- iii) comparing the plurality of images in each of the plurality of first identification groups and thereby counting a number of different license plates in each of the plurality of first identification groups,
- b) an identification process, used subsequent to the learning process, the identification process comprising:
- i) acquiring a second image of the target license plate, and,
- ii) identifying, solely by optical character recognition, a character string on the target license plate in the second image of the target license plate, and comparing the character string on the second image of the target license plate with the character strings associated with each of the plurality of first identification groups, and,
- iii) selecting the first identification group whose associated character string is the same as the character string of the second image of the target license plate, producing a selected first identification group, and,
- iv) if the number of different license plates in the selected first identification group is one, identify the characters on the target license plate as the character string associated with the selected first identification group, and,
- if the number of different license plates in the selected first identification group is greater than one, comparing pairwise each of the plurality of images in the selected first identification group with the second image of the target licensed plate, the comparison providing a confidence estimate for a match of each pairwise compared images, and,
- v) identify the characters on the target license plate as the character string associated with the image in the selected first identification group that has the highest confidence estimate.
Dependencies are inferred from the claim text.