System, computing device, and method for document detection and deposit processing
View Patent ↗An image of a check is captured by an imaging device and processing of the digital image of the check for deposit at a remote server may be accomplished with a downloaded software application on a portable computing device associated with the imaging device. The downloaded application may include one or more trained machine learning models for processing the captured image. The portable computing device may utilize deterministic algorithms for certain image processing tasks and machine learning models for others. The selection between machine learning and deterministic processing may be made locally on the portable device or in response to instructions from an institution server to use a particular processing method.
1. A portable computing device, comprising:
a processor;
memory; and
instructions stored on the computer-readable medium, the instructions configured to, when executed by the processor, cause the processor to:
capture a digital check image depicting a check;
input the data from the captured digital check image into a trained machine learning image cropping model to determine a plurality of edge coordinates corresponding to the check depicted in the digital check image;
detect a check image included in the digital check image based on the plurality of edge coordinates;
generate a digital image file including the digital check image and the edge coordinates; and
transmit the digital image file to a remote server for deposit.
2. The computing device of claim 1 , wherein the instructions are configured to, when executed by the processor, further cause the processor to:
remove the background image from the digital check image by cropping the digital check image based on the edge coordinates determined by the trained machine learning model.
3. The computing device of claim 1 , wherein the instructions are configured to, when executed by the processor, further cause the processor to generate the digital image file to include check data received via user input.
4. The computing device of claim 1 , wherein the instructions are configured to, when executed by the processor, further cause the processor to:
input the data from the captured digital check image into a trained machine learning image denoising model;
remove handwriting from printed text on the check depicted in the digital check image based on an output of the trained machine learning image denoising model.
5. The computing device of claim 1 , wherein the instructions are configured to, when executed by the processor, further cause the processor to:
input the data from the captured digital check image into a trained machine learning image dewarping model;
obtain a coordinate conversion map for dewarping the check depicted in the digital check image based on an output of the trained machine learning image dewarping model.
6. The computing device of claim 5 , wherein the instructions are configured to, when executed by the processor, further cause the processor to:
input the data from the captured digital check image into a trained machine learning duplicate model, wherein the trained machine learning duplicate check model comprises a perceptual hash trained model;
when the trained machine learning duplicate model is stored on the portable computing device, prevent transmission of the data from the captured check to the remote server in response to a determination by the trained machine learning duplicate model that the check depicted in the digital check image based is a duplicate check.
7. The computing device of claim 4 , wherein the instructions are configured to, when executed by the processor, further cause the processor to:
input the data from the captured digital check image, after removing the handwriting from the printed text based on an output of the trained machine learning image denoising model, into a trained machine learning image dewarping model;
obtain a coordinate conversion map for dewarping the check depicted in the digital check image based on an output of the trained machine learning image dewarping model.
8. A portable computing device, comprising:
a processor;
a memory; and
wherein the processor is configured to:
download an app for performing image processing on, and remotely depositing to an institution, a digital check image of a check, wherein the downloaded app comprises a trained machine learning model for executing a first image processing task:
control, using the downloaded app, an image capture device associated with the portable computing device to capture the digital check image;
input, using the downloaded app, the captured digital check image into the trained machine learning model to modify the digital check image in accordance with the first image processing task;
process the modified digital check image, using the downloaded app, with a deterministic algorithm to execute a second image processing task that is different from the first image processing task;
transmit the modified and processed digital check file to a remote server for deposit.
9. The portable computing device of claim 8 , wherein the trained machine learning model for executing the first image processing task comprises an image cropping model configured to cause the processor of the portable computing device to remove a background image from the digital check image by cropping the digital check image based on edge coordinates determined by the trained machine learning model.
10. The portable computing device of claim 8 , wherein the trained machine learning model for executing the first image processing task comprises an image demorphing model configured to cause the processor of the portable computing device to remove a skew from the digital check image.
11. The portable computing device of claim 10 , wherein the image demorphing model causes the processor to generate a spatial transformation matrix to apply to the digital check image.
12. The portable computing device of claim 8 , wherein the trained machine learning model for executing the first image processing task comprises a denoising model configured to cause the processor of the portable computing device to remove handwritten marks from overlapping printed text in the digital check image.
13. The portable computing device of claim 8 , wherein the portable computing device comprises a smart phone and image capture device associated with the portable computing device comprises a digital camera.
14. The portable computing device of claim 8 , wherein the downloaded app comprises a plurality of trained machine learning models and the processor is configured by the downloaded app to select between two or more trained machine learning models for use in a particular image processing task after determining a state of the captured digital check image.
15. A portable computing device, comprising:
a processor;
a memory; and
wherein the processor is configured to:
download an app for performing image processing on, and remotely depositing to an institution, a digital check image of a check, wherein the downloaded app comprises a first trained machine learning model;
control, using the downloaded app, an image capture device associated with the portable computing device to capture the digital check image;
input, using the downloaded app, the captured digital check image into the first trained machine learning model to identify which of a plurality of second trained machine learning models is a best match for handling an image processing task; and
based on the determined best match for handling the image processing task, select and execute a second trained machine learning model to modify the captured digital check image.
16. The portable computing device of claim 15 , wherein the first trained machine learning model is trained to cause the processor to:
recognize one of a specific check type or check design type of the captured digital check image; and
select one of the plurality of second trained machine learning models that is a best match for handling the image processing task based on the check type or check design type.
17. The portable computing device of claim 16 , wherein the first trained machine learning model is further trained to cause the processor to:
recognize at least one environmental factor associated with the captured digital check image, wherein the environmental factor comprises one of a lighting or shadow detection in the captured digital check image; and
cause the processor to select the one of the plurality of second trained machine learning models based on the recognized environmental factor.
18. The portable computing device of claim 16 , wherein the processor is configured to, when the second trained machine learning model determined as the best match for handling the image processing task is not stored in the memory on the portable computing device, download the second trained machine learning model from a remote computer.
19. The portable computing device of claim 15 , wherein the first trained machine learning model is further configured to cause the processor to alter a device setting of the portable computing device for an capture operation based on a currently detected environmental condition.
20. The portable computing device of claim 15 , wherein the first trained machine learning model is further configured to cause the processor to alter a default device setting of the portable computing device for a future image capture operation after a current image capture operation based on a currently detected environmental condition.