Image Processing Apparatus for Regularly Arranged Product Inspection
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing image processing methods struggle with efficiently processing a plurality of works arranged regularly, leading to difficulties in determining the number of products packed in one package, long setup times, and potential production yield losses due to failed automatic division of inspection areas with repetitive patterns.
Innovation Solution
An image processing apparatus that defines and processes multiple target areas on an input image, allowing for activation or deactivation of processing for each area, with neighboring areas arranged in a matrix, zigzag, or radial patterns, using a single pre-registered model for pattern matching, and enabling manual setting of process target areas for efficient inspection.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If pattern matching is executed on the entire input image containing multiple regularly arranged works, then all works can be detected in one process, but multiple positions of the same work are incorrectly detected as matching the model, making it difficult to determine the actual number of products
Solution Approach 1:
The input image is divided into multiple process target areas corresponding to the regular arrangement of works. By defining specific search ranges for each work position and executing pattern matching independently in each area, the system avoids detecting the same work multiple times and accurately determines the actual number of products in the package.
2Measurement precision
If multiple models are independently set to detect the number of works in one input image, then accurate detection is possible, but the setting procedure takes much time
Solution Approach 1:
A single pre-registered model is used for pattern matching across all process target areas. The model serves multiple functions by being applied repeatedly to different areas of the input image, eliminating the need to create and set multiple independent models while maintaining accurate work detection.
3Extent of automation
If automatic division of inspection areas is used for images with repetitive patterns, then the process is automated, but the division may fail causing measurement process to stop and lowering production yield
Solution Approach 1:
Process target areas are defined in advance based on the known regular arrangement of works, including the number of works, their positions, and search ranges. This preliminary definition prevents automatic division failures by establishing clear boundaries before pattern matching begins, ensuring continuous and reliable measurement processing.
4Loss of time
If a single model is used for pattern matching across all areas, then setup time is reduced, but it becomes difficult to handle different inspection requirements for different areas
Solution Approach 1:
While using a single model for pattern matching, the system allows dynamic activation or inactivation of individual process target areas based on inspection requirements. This enables flexible handling of different inspection scenarios by selectively enabling or disabling specific areas without changing the model, maintaining both efficiency and adaptability.
Data Source
Figure 1
Figure 2
Figure 3
AI summary
An image processing apparatus (100) in accordance with an embodiment receives a setting related to a common image processing to be executed on each of a plurality of process target areas (S 11); receives a setting of a reference area for defining the plurality of process target areas on an input image (S12); and receives a setting for regularly defining the plurality of process target areas using the reference area as a reference (S13). In accordance with the setting related to the common image processing, image processing is executed on each of the plurality of process target areas, and a result of overall process reflecting the results of image processing of respective ones of the plurality of process target areas is output (S26).