Component Mounting Machine Image Classification via Multi-Camera Verification
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
In component mounting machines, images captured by cameras may incorrectly classify components due to positional deviations caused by waste or dust, leading to incorrect calculation and storage of images as normal, even when mounting errors occur.
Innovation Solution
The system classifies images based on current and prior imaging timings, determining normality and abnormality to prevent incorrect storage of abnormal images as normal, using multiple cameras to capture images at different timings during the mounting process and incorporating a control device to manage and correct positional deviations.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If images are classified based solely on positional deviation within allowable range, then classification speed is improved, but image classification accuracy deteriorates due to incorrect normal/abnormal judgment
Solution Approach 1:
The system performs preliminary classification based on positional deviation to quickly identify obviously normal images, then applies secondary verification using multiple camera views to confirm the classification before final storage. This preliminary action approach maintains high processing speed while adding accuracy verification steps.
Solution Approach 2:
The system uses feedback from multiple camera perspectives (side camera and part camera) to verify the initial classification result. If there is any doubt or inconsistency in the preliminary classification, the feedback mechanism triggers re-evaluation using additional image data to ensure accurate normal/abnormal judgment.
2Reliability
If multiple cameras are used to capture images at multiple timings, then image classification accuracy is improved, but device complexity increases
Solution Approach 1:
The multiple cameras (side camera and part camera) serve universal functions: they capture images at different angles and timings, provide redundant verification data, and can identify various types of abnormalities (positional deviation, mounting errors, component defects). This multi-functionality approach improves classification accuracy without proportionally increasing system complexity.
Solution Approach 2:
The imaging process is segmented into multiple stages: initial imaging by the side camera, intermediate verification by the part camera, and final classification. Each camera segment performs a specific function in the overall classification workflow, allowing the system to manage complexity through functional segmentation while maintaining high accuracy.
3Productivity
If images are stored without verification of mounting error, then storage efficiency is improved, but data reliability deteriorates due to incorrect normal image storage
Solution Approach 1:
The system performs preliminary verification by checking mounting error status before final image storage. This preliminary action ensures that only confirmed normal images are stored in the normal image database, preventing contamination of the database with abnormal images while maintaining efficient storage operations.
Solution Approach 2:
The control device acts as an intermediary between image capture and image storage, verifying mounting error status and classification accuracy before allowing images to be stored. This intermediary verification step ensures data reliability without significantly impacting storage efficiency, as it only filters out clearly abnormal cases.
Data Source
Figure 1
Figure 2
Figure 3
AI summary
A component mounting machine includes a component supply device configured to supply a component, a head having a nozzle, a movement device configured to move the head in XY-directions, and a control device configured to control each device. The control device is configured to control each device such that a series of works including a pickup operation of picking up the component by the nozzle, a movement operation of moving the head, and a mounting operation of mounting component picked up by the nozzle on a substrate is executed. The control device is configured to control a cameras to image the component when an imaging timing reaches, determine the normality and abnormality of an operation executed immediately before a current imaging timing based on an image captured at the current imaging timing, classify the image captured at the current imaging timing as a normal image in a case in which a result of determination of the normality and abnormality is made as normal, and classify the image captured at the current imaging timing and a part or all of images captured at imaging timings prior to the current imaging timing as an abnormal image in a case in which the result of the determination of the normality and abnormality is made as abnormal.