Packaging Seal Inspection Training Data via Image Region Cutout
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing methods lack an efficient approach for generating training data for learned models used in inspecting the sealing sections of packaged items to determine non-defective or defective status.
Innovation Solution
A training data generation device and inspection device that utilize a cut-out means to specify and extract inspection areas based on reference points or blob analysis, associate images with sorting results, and generate training data for machine learning models.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If learned models are used as determination criteria for inspection, then inspection accuracy for defects like biting and sealing issues is improved, but the complexity of data collection and training data generation increases
Solution Approach 1:
The patent segments the image processing into distinct stages: receiving full images, specifying inspection areas, cutting out regions of interest, and generating training data. This segmentation simplifies the overall complex task by breaking it into manageable steps that can be automated and standardized.
Solution Approach 2:
The patent performs preliminary actions by automatically specifying inspection areas and cutting out regions of interest before training data generation. This preliminary processing prepares the data in advance, reducing the complexity of subsequent training data collection and enabling more efficient learned model development.
2Reliability
If training data is generated manually for machine learning, then data accuracy is improved, but time consumption and productivity decrease
Solution Approach 1:
The system performs self-service by automatically specifying inspection areas, cutting out regions of interest, and generating training data without manual intervention. The automated inspection area specification and image cutting processes enable the system to prepare accurate training data independently, significantly improving productivity while maintaining reliability through consistent automated processing.
Solution Approach 2:
The patent replaces manual mechanical data preparation processes with automated computational methods. Instead of manually selecting and processing images, the system uses automated image processing algorithms to specify inspection areas, cut out regions, and generate training data, thereby increasing efficiency while maintaining accuracy through systematic automated processing.
3Reliability
If entire images are used for inspection, then comprehensive defect detection is improved, but processing time and computational load increase
Solution Approach 1:
The patent extracts only the relevant inspection areas from full images by automatically specifying and cutting out regions of interest. This extraction process removes unnecessary portions of images, reducing processing time and computational load while maintaining comprehensive defect detection within the extracted inspection areas.
Solution Approach 2:
The patent segments full images into smaller inspection areas or regions of interest based on automatically specified coordinates. This segmentation reduces the amount of data that needs to be processed while ensuring that all relevant defect areas are covered, thereby decreasing processing time without sacrificing detection comprehensiveness.
Data Source
AI summary
A training data generation device includes: cut-out unit that receives an input of an image of an inspection target, specifies an inspection area of the inspection target in the image by a predetermined method, and cuts out the specified inspection area from the image; sorting unit that, on the basis of a sorting operation of sorting, as a learning-target image, an image of the inspection area cut out in the cut-out unit into at least either normal or non-normal, associates the learning-target image and a result of the sorting with each other; and training data memory unit that stores training data in which the learning-target image and the result of the sorting are associated with each other.