Machine Learning Check Image Detection for Remote Deposit
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Traditional deterministic algorithms for remote check deposit systems face challenges in accurately detecting check images, especially with insufficient contrast, reflections, shadows, and distortions, leading to delays or failures in processing.
Innovation Solution
The implementation of machine learning models trained to identify check images from backgrounds, de-warp distorted images, and remove noise, enabling more robust edge detection and duplicate check detection across varying conditions.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If deterministic algorithms are used to analyze check images, then the processing is straightforward and deterministic, but the accuracy decreases when images have varying characteristics such as insufficient contrast, reflections, shadows, and distortions
Solution Approach 1:
The patent replaces deterministic mechanical algorithms with machine learning models that can adaptively learn and recognize check image features. The machine learning models are trained on diverse check images with various characteristics including different contrasts, reflections, shadows, and distortions, enabling them to generalize and accurately detect check images under varying conditions without requiring explicit programming for each scenario.
2Reliability
If machine learning models are implemented to handle diverse image conditions, then the accuracy and reliability improve, but the system complexity increases
Solution Approach 1:
The patent applies preliminary action by pre-training machine learning models on extensive datasets of check images with various characteristics before deployment. This pre-training phase allows the models to learn robust feature representations and detection patterns in advance, so that during actual remote check deposit processing, the models can directly apply their learned knowledge without requiring complex real-time adjustments or additional processing complexity.
3Productivity
If traditional algorithms are used, then the system is simpler to implement, but processing delays or failures occur with diverse image conditions
Solution Approach 1:
The patent employs parameter changes by utilizing machine learning models that can dynamically adapt their internal parameters and decision boundaries based on the characteristics of the input image. Unlike fixed deterministic algorithms, the machine learning models adjust their processing parameters according to the specific conditions of each check image, such as contrast levels, lighting conditions, and distortion types, thereby maintaining high detection success rates across diverse image conditions without sacrificing processing efficiency.
Data Source
AI summary
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.