Acceptance Audit Image Normalization for False Order Detection
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Traditional procurement acceptance audits are slow and inefficient, particularly in detecting abnormalities in images within procurement acceptance transaction documents, due to manual inspection methods that cannot verify all documents effectively.
Innovation Solution
A computer-based method that processes image data by counting height-width categories, adjusting image dimensions, performing image augmentation, and calculating similarity indices to identify similar or non-similar images, thereby automating the detection of potential false acceptances or falsified orders.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If manual inspection methods are used for procurement acceptance audits, then inspection accuracy can be maintained through human judgment, but audit speed and productivity are significantly reduced
Solution Approach 1:
The patent replaces manual mechanical inspection with an automated image processing system that uses computer vision algorithms to analyze procurement acceptance images. The system automatically detects abnormalities such as duplicate images, inconsistent product quantities, and documentation mismatches, achieving both high-speed processing and maintained inspection accuracy through technological substitution rather than human judgment.
Solution Approach 2:
The patent creates digital copies of procurement acceptance images and processes these copies through automated analysis algorithms. By working with digital reproductions rather than physical documents, the system can rapidly analyze multiple images simultaneously while maintaining the ability to detect subtle abnormalities that would require careful human examination, thus resolving the speed-accuracy tradeoff.
2Reliability
If all procurement acceptance transaction documents are verified, then detection of abnormalities is improved, but the time and resources required for complete verification increase significantly
Solution Approach 1:
The patent implements a multi-layered verification approach where the image processing system performs rapid preliminary analysis on all images to identify obvious abnormalities, then applies more intensive analysis only to suspicious cases. This partial exhaustive approach (analyzing all images but with varying depths of scrutiny) maintains high reliability for detecting abnormalities while significantly reducing the total time required compared to uniform detailed verification of every document.
Solution Approach 2:
The system performs preliminary automated filtering and classification of images before detailed inspection. By pre-processing images to identify potential issues such as duplicate hashes, unusual patterns, or metadata inconsistencies, the system prepares a prioritized list for deeper analysis, thereby achieving comprehensive abnormality detection with reduced overall verification time through advance preparation.
3Productivity
If image dimensions are standardized for processing, then processing efficiency is improved, but images with non-standard dimensions require additional adjustment steps
Solution Approach 1:
The patent implements automatic parameter adaptation where the image processing system dynamically adjusts processing parameters based on the actual dimensions and characteristics of each input image. Rather than requiring all images to conform to fixed standards, the system modifies its processing pipeline to accommodate various image formats, resolutions, and aspect ratios, thereby maintaining high processing efficiency without adding manual adjustment complexity for non-standard images.
Data Source
AI summary
A method for acceptance inspection audit is implemented by a computer that stores original image data sets. Each original image data set corresponds to an object image, and an image height-width data set of the object image that corresponds to one of image height-width categories. The method includes steps of: A) based on the image height-width data sets, counting a number of occurrences of each of the image height-width categories; B) obtaining, as a target image height-width category, one of the image height-width categories that has a greatest number of occurrences; C) for each original image data set in a group of to-be-inspected data sets, adjusting the object image based on the target image height-width category, thereby obtaining an adjusted-original image data set; and D) obtaining at least one image inspection result based at least on the adjusted-original image data sets thus obtained respectively for the original image data sets.