Project:
The customer has had already implemented a solution witch uses classical object recognition to identify, but he was not satisfied with the results.
Problems:
- There was not reliable accuracy of quality control.
- Because of high standards in his industry.
Our Solutions
We provided a function prototype solution to soldering control by vision system using deep neural networks. Today, a system designed in this way represents a fast and efficient method of defectoscopy.
Soldered bulbs come out of the automated line and pass a functionality test. When these bulbs are fitted to primary production, a number of strict quality criteria need to be met, which are not purely related to functionality. For this reason, it is necessary to perform a final inspection of the quality of the bulb, where the classical methods fail to meet these high criteria in the required quality.
Benefits and results:
- Reliable solution with more than 90% accuracy
- Ability to operate 24/7
- Higher quality of output product
