Empirical method and system for increasing the attach rate of OCR engines to images
Methods and systems for automatic license plate recognition (ALPR) can involve obtaining results from a manual image review of an image captured by one or more ALPR cameras and processed by an ALPR engine and feeding the results from the manual image review back to the ALPR engine to train the ALPR engine where to search in the image for a license plate. Training of the ALPR engine increase the attach rate of an optical character recognition (OCR) engine with respect to images captured the ALPR camera(s) and processed by the ALPR engine.
1. A method for automatic license plate recognition (ALPR), comprising:
extracting results from a manual image review of an image among a plurality of images captured by at least one ALPR camera and processed by an ALPR engine associated with an OCR engine and which includes the at least one ALPR camera, wherein the manual image review occurs after an initial automatic license plate review of the image;
feeding the results from the manual image review in a processing pipeline back to the ALPR engine to train the ALPR engine; and
training the ALPR engine where to search in a region of interest of the image for a license plate after feeding the results from the manual image review in the processing pipeline back to the ALPR engine to facilitate by the training of the ALPR engine an increase in an attach rate of the OCR engine to the plurality of images.
2. The method of claim 1 , further comprising identifying the license plate in the image.
3. The method of claim 1 further comprising automatically magnifying a region of interest in the image after extracting the results from the manual image review of the image processed by the ALPR engine.
4. The method of claim 1 wherein the results from the manual image review are fed back to the ALPR engine after automatically magnifying the region of interest in the image.
5. The method of claim 1 further comprising outputting from the OCR engine a confidence score for each of a segmented character from which an overall confidence score is calculated for all of the license plate, wherein if the overall confidence score is higher than a pre-defined threshold, a recognized license plate number of the license plate is transmitted to a remaining portion of the processing pipeline without human interruption.
6. The method of claim 1 further comprising:
automatically magnifying a region of interest in the image after extracting the results from the manual image review of the image processed by the ALPR engine, wherein the results from the manual image review are fed back to the ALPR engine after automatically magnifying the region of interest in the image.
7. The method of claim 1 wherein the ALPR engine further comprises parking software and video management software.
8. A system for automatic license plate recognition (ALPR), comprising:
at least one ALPR camera that captures an image; and
an ALPR engine comprising the at least one ALPR camera, wherein the ALPR engine is associated with an OCR engine and communicates with the at least one ALPR camera, wherein results are extracted from a manual image review of the image among a plurality of images captured by the at least one ALPR camera and processed by the ALPR engine, and wherein the results are fed in a processing pipeline from the manual image review back to the ALPR engine to train the ALPR engine where to search in a region of interest of the image for data indicative of a license plate after feeding the results from the manual image review in the processing pipeline back to the ALPR engine to facilitate by the training of the ALPR engine an increase in an attach rate of the OCR engine to the plurality of images, wherein the manual image review occurs after an initial automatic license plate review of the image.
9. The system of claim 8 , wherein the ALPR engine identifies the license plate in the image based on the data indicative of the license plate.
10. The system of claim 9 wherein the ALPR engine automatically magnifies the region of interest in the image after extracting the results from the manual image review of the image processed by the ALPR engine.
11. The system of claim 9 wherein the results from the manual image review are fed back to the ALPR engine after automatically magnifying the region of interest in the image.
12. The system of claim 9 wherein: a confidence score is output from the OCR engine for each of a segmented character from which an overall confidence score is calculated for all of the license plate, wherein if the overall confidence score is higher than a pre-defined threshold, a recognized license plate number of the license plate is transmitted to a remaining portion of the processing pipeline without human interruption.
13. The system of claim 9 wherein:
the ALPR engine automatically magnifies the region of interest in the image after obtaining the results from the manual image review of the image processed by the ALPR engine, wherein the results from the manual image review are fed back to the ALPR engine after automatically magnifying the region of interest in the image.
14. The system of claim 9 wherein:
the ALPR engine automatically magnifies the region of interest in the image after obtaining the results from the manual image review of the image processed by the ALPR engine.
15. A non-transitory computer program product for automatic license plate recognition (ALPR), the non-transitory computer program product comprising one or more computer readable storage media and program instructions collectively stored on the one or more computer readable storage media, the program instructions comprising program instructions to:
extract results from a manual image review of an image among a plurality of images captured by at least one ALPR camera and processed by an ALPR engine associated with an OCR engine and which includes the at least one ALPR camera, wherein the manual image review occurs after an initial automatic license plate review of the image;
feed the results from the manual image review back to the ALPR engine to train the ALPR engine; and
train the ALPR engine where to search in a region of interest of the image for a license plate after feeding the results from the manual image review in the processing pipeline back to the ALPR engine to facilitate by the training of the ALPR engine an increase in an attach rate of the OCR engine to the plurality of images.
16. The non-transitory computer program product of claim 15 , wherein the program instructions further comprise program instructions to identify the license plate in the image.
17. The non-transitory computer program product of claim 15 , wherein the program instructions further comprise program instructions to automatically magnify the region of interest in the image after obtaining the results from the manual image review of the image processed by the ALPR engine.
18. The non-transitory computer program product of claim 15 , wherein the results from the manual image review are fed back to the ALPR engine after automatically magnifying the region of interest in the image.
19. The non-transitory computer program product of claim 15 , wherein the program instructions further comprise program instructions to output a confidence score from the OCR engine for each of a segmented character from which an overall confidence score is calculated for all of the license plate, wherein if the overall confidence score is higher than a pre-defined threshold, a recognized license plate number of the license plate is transmitted to a remaining portion of the processing pipeline without human interruption.
20. The non-transitory computer program product of claim 15 , wherein the program instructions further comprise program instructions to automatically magnify the region of interest in the image after obtaining the results from the manual image review of the image processed by the ALPR engine, wherein the results from the manual image review are fed back to the ALPR engine after automatically magnifying the region of interest in the image.