Method and system for container code recognition

1. A method for determining container identification code via Spatial Transformer Networks (STN) and Connected Component (CC), said method comprising a hardware processor implemented steps of:

capturing an image of a container using a camera, wherein the image contains the container identification code;

pre-processing the captured image;

extracting and filtering region proposals from the pre-processed image to generate regrouped region proposals, wherein extraction and filtration is performed by implementing connected components (CC);

classifying the regrouped region proposals into characters by implementing trained Spatial Transformation Network (STN) to generate a valid group of region proposals with more than one chunk of container identification code; and

generating a sequence for the valid group of region proposals and mapping the generated sequence to a predefined standard container identification code to determine the container identification code, wherein the predefined standard identification code comprises chunks of characters in a predefined pattern, wherein generating the sequence comprising:

determining the container identification code by matching the sequence of valid region proposals to the predefined standard container code when all chunks of characters of the predefined standard identification code are contained in at least one sequence of the valid region proposal; and

identifying, using heuristics, remaining chunks of characters in a sequence of valid region proposals when all chunks of characters of the predefined standard identification code are not contained in at least one sequence of the valid region proposal and determining the container identification code by matching the sequence of valid region proposals and generated remaining chunks to the predefined standard identification code.

2. The method according to claim 1 further comprising triggering an alert when the container identification code is not determined after generating and mapping the valid group of region proposals.

3. The method according to claim 1 wherein the container identification code is an ISO (International Standard Organization) 6346 code, further comprising, validating the determined container identification code using checksum digit for ISO 6346 code.

4. The method according to claim 1 wherein image preprocessing comprises the steps of:

resizing the images to double the original size of the image; and

binarizing the resized image separate one or more characters of the container code in the image.

5. The method according to claim 1 wherein generating the regrouped region proposals comprises the steps of:

extracting the region proposals using connected components wherein the extracted region proposals are according to the structure of the predefined standard code;

filtering false positive from the extracted region proposals; and

regrouping the region proposals after elimination of false positives based on similarity of spatial position and sequence of region proposals of similar heights.

6. The method according to claim 1 wherein generate a valid group of region proposals comprises the steps of:

classifying the grouped region proposals to characters by using trained STNs including a STN for alphabets (STNalp) and a STN for digits (STNd); and

matching the grouped region proposals to the characters from a predefined standard code directory to generate valid region proposals wherein the predefined standard code directory comprises at least one chunk of characters of the predefined standard identification code.

7. A system ( 102 ) for determining container identification code via Spatial Transformer Networks (STN) and Connected Component (CC), comprising a processor ( 202 ), a memory ( 204 ) and a camera operatively coupled with said processor, the system configured to:

capture an image of a container wherein the image contains the container identification code;

pre-process the captured image;

extract and filter region proposals from the pre-processed image to generate regrouped region proposals, wherein extraction and filtration is performed by implementing connected components (CC);

classify the regrouped region proposals into characters by implementing trained Spatial Transformation Network (STN) to generate a valid group of region proposals with more than one chunk of container identification code; and

generate a sequence for the valid group of region proposals and mapping the generated sequence to a predefined standard container identification code to determine the container identification code, wherein the predefined standard identification code comprises chunks of characters in a predefined pattern, wherein the system is further configured to:

determine the container identification code by matching the sequence of valid region proposals to the predefined standard container code when all chunks of characters of the predefined standard identification code are contained in at least one sequence of the valid region proposal; and

identify, using heuristics, remaining chunks of characters in a sequence of valid region proposals when all chunks of characters of the predefined standard identification code are not contained in at least one sequence of the valid region proposal and determining the container identification code by matching the sequence of valid region proposals and generated remaining chunks to the predefined standard identification code.

8. The system according to claim 7 is further configured to trigger an alert when a container identification code is not determined after generating and mapping the valid group of region proposals.

9. The system according to claim 7 is further configured to:

resize the images to double the original size of the image; and

binarize the resized image separate one or more characters of the container code in the image.

10. The system according to claim 7 is further configured to:

extract the region proposals using connected components (CC) wherein the extracted region proposals are according to the structure of the predefined standard code;

filter false positive from the extracted region proposals; and

regroup the region proposals after elimination of false positives based on similarity of spatial position and sequence of region proposals of similar heights.

11. The system according to claim 7 is further configured to:

classify the grouped region proposals to characters by using trained STNs including a STN for alphabets (STNalp) and a STN for digits (STNd); and

match the grouped region proposals to the characters from a predefined standard code directory to generate valid region proposals wherein the predefined standard code directory comprises at least one chunk of characters of the predefined standard identification code.

12. One or more non-transitory machine readable information storage mediums comprising one or more instructions which when executed by one or more hardware processors causes the one or more hardware processors to perform a method for determining container identification code via Spatial Transformer Networks (STN) and Connected Component (CC), said method comprising:

capturing an image of a container using a camera, wherein the image contains the container identification code;

pre-processing the captured image;

extracting and filtering region proposals from the pre-processed image to generate regrouped region proposals, wherein extraction and filtration is performed by implementing connected components (CC);

classifying the regrouped region proposals into characters by implementing trained Spatial Transformation Network (STN) to generate a valid group of region proposals with more than one chunk of container identification code; and

generating a sequence for the valid group of region proposals and mapping the generated sequence to a predefined standard container identification code to determine the container identification code, wherein the predefined standard identification code comprises chunks of characters in a predefined pattern, wherein generating the sequence comprising:

determining the container identification code by matching the sequence of valid region proposals to the predefined standard container code when all chunks of characters of the predefined standard identification code are contained in at least one sequence of the valid region proposal; and

identifying, using heuristics, remaining chunks of characters in a sequence of valid region proposals when all chunks of characters of the predefined standard identification code are not contained in at least one sequence of the valid region proposal and determining the container identification code by matching the sequence of valid region proposals and generated remaining chunks to the predefined standard identification code.

13. The one or more non-transitory machine readable information storage mediums of claim 12 , further comprising triggering an alert when the container identification code is not determined after generating and mapping the valid group of region proposals.

14. The one or more non-transitory machine readable information storage mediums of claim 12 , wherein the container identification code is an ISO (International Standard Organization) 6346 code, further comprising, validating the determined container identification code using checksum digit for ISO 6346 code.

15. The one or more non-transitory machine readable information storage mediums of claim 12 , wherein image preprocessing comprises the steps of:

resizing the images to double the original size of the image; and

binarizing the resized image separate one or more characters of the container code in the image.

16. The one or more non-transitory machine readable information storage mediums of claim 12 , wherein generating regrouped region proposals comprises the steps of:

extracting the region proposals using connected components wherein the extracted region proposals are according to the structure of the predefined standard code;

filtering false positive from the extracted region proposals; and

regrouping the region proposals after elimination of false positives based on similarity of spatial position and sequence of region proposals of similar heights.

17. The one or more non-transitory machine readable information storage mediums of claim 12 , wherein generate a valid group of region proposals comprises the steps of:

classifying the grouped region proposals to characters by using trained STNs including a STN for alphabets (STNalp) and a STN for digits (STNd); and

matching the grouped region proposals to the characters from a predefined standard code directory to generate valid region proposals wherein the predefined standard code directory comprises at least one chunk of characters of the predefined standard identification code.