Method and system for analyzing a machine vision image
By stitching and clustering training datasets based on product and business information from machine-inspected images, confidence intervals are generated, solving the problem of wasted machine-inspected image resources and achieving efficient risk analysis and accurate anomaly detection.
CN122289831APending Publication Date: 2026-06-26NUCTECH JIANGSU CO LTD +1
View PDF 0 Cites 0 Cited by
Patent Information
- Authority / Receiving Office
- CN ยท China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- NUCTECH JIANGSU CO LTD
- Filing Date
- 2024-12-26
- Publication Date
- 2026-06-26
AI Technical Summary
Technical Problem
The massive amount of machine-generated images occupies a large amount of storage space and is not fully utilized, resulting in serious waste of resources. Existing technologies lack effective intelligent analysis methods.
Method used
By acquiring unseen anomaly data corresponding to product information and business information as a training dataset, performing splicing and clustering analysis, training a discrete analysis model, and generating confidence intervals for risk assessment.
Benefits of technology
Effective management of training models improves analytical accuracy, integrates different feature information, and enables efficient utilization and risk analysis of machine inspection data.
โฆ Generated by Eureka AI based on patent content.
Smart Images
Figure CN122289831A_ABST
Abstract
This invention relates to a method and system for analyzing machine-inspected images. The method includes: acquiring no-anomaly data corresponding to each strategy as a training dataset, wherein the no-anomaly data includes machine-inspected data where the machine inspection conclusion for the images is no anomaly, and the strategy is set in correspondence with product information and / or business information; performing the following stitching process on the training dataset: if multiple image features are selected from each machine-inspected image in the training dataset, the multiple image features are stitched together; if only one image feature is selected from each machine-inspected image, no stitching is performed; and training a discrete analysis model using the stitched training dataset and saving it in correspondence with the strategy. According to this application, historical data can be utilized, the training model can be effectively managed, thereby improving the accuracy of the analysis.
Need to check novelty before this filing date? Find Prior Art