Unlike manual visual inspection methods, which are often subjective, you get repeatable and accurate results.
Changing image conditions or high error variations are not a problem compared to classical methods.
Significantly shorter evaluation times than with classic image processing allow automated inline operation.
Project Subtitle
System for adaptive phototonic surface testing with adaptive image evaluation in combination with a cleaning system.
Companies and Partners
P…
Project Subtitle
Automation of network edge infrastructure & applications with artificial intelligence
Companies & Partners
Opel Automobile GmbH, Frau…
Project Subtitle
Data-based evaluation of the wire electrical discharge machining process
Companies & Partners
WBA Aachener Werkzeugbau Akademie GmbH, Mak…
Understand use-case and quality evaluation criteria
Already image data available? Know its suitability
Collect & structure meaningful image data
Know the return on investment considering costs and risks
Plan shopfloor deployment & data flow with your IT
Get a roll-out roadmap after proof of concept
Learn more about AI in Manufacturing and its use-cases.
What is Machine Vision?
Machine Vision is a term that describes the process of using digital cameras, sensors, and processing algorithms to capture and interpret visual information from the physical world. This technology is used in a variety of applications, including industrial robotics, medical imaging, and autonomous vehicles. It enables machines to “see” and make decisions in real time, allowing them to respond to their environment in an intelligent and efficient manner.
How does Machine Vision work for Visual Quality Control?
In visual quality control, computer-based solutions for task profiles are typically more likely to match the capabilities of the human visual system. In this context, Machine Vision methods are considered a tool of choice, using artificial neural networks with a large number of layers, so-called hidden layers. Prerequisites are usually a sufficient amount of training data as well as corresponding computing capacities of the hardware used for training and evaluation. Alternatively, in some cases, pre-trained algorithms can be used. In this case, highly complex problems can be solved for which no or only very complex robust solutions can be developed with classical rule-based algorithms.
How is Machine Vision used for Visual Quality Control?
Today, more and more industrial image processing systems are used for quality assurance. These systems can be installed on the production line in series production and can reliably and repeatably inspect thousands of parts per minute in some cases. Even in the production of smaller quantities, Machine Vision offers excellent results in quantitative measurements of structured scenes due to the accuracy and repeatability achieved. In combination with the appropriate resolution and optics, details can be detected that are invisible or difficult to see for the human eye. Machine Vision inspection thus often replaces human inspection processes, is objective and repeatable, and potentially carries fewer errors and failures.
What are the benefits of Machine Vision for Visual Quality Control?
The core problem with regard to conventional Machine Vision systems is that they cannot distinguish between “external” variations in the image data and the functional or qualitative deviations of the inspection objects that are relevant for Visual Quality Control.
In contrast to classical image processing, humans are able to learn and make these distinctions. Even though humans process information more slowly, their strength lies in being able to abstract it and evaluate it against a more general background. Therefore, in use cases where qualitative interpretation of complex, unstructured, i.e. significantly varying, image data is required, human inspection is generally the better choice.
With the help of Machine Vision models, cognitive human capabilities can be partially transferred to Machine Vision systems. In this way, the abstracting, learning and qualitative interpretation of image data with its robustness to variance can be combined with the speed and repeatability of industrial Machine Vision systems.
What should be considered before using Machine Vision for Visual Quality Control?
Machine vision methods for image processing are particularly well suited for complicated inspection procedures in production, which must be robust to variations or anomalies in the image data and evaluate qualitative or discrete measured variables.
Artificial neural networks enable the automation of processes for which it was previously impossible or very difficult to find programmable solutions. In this way, error rates can be minimized and inspection times shortened. The entry into the implementation of Machine Vision applications in image processing is facilitated by existing programming libraries such as TensorFlow or even already completely prefabricated software and requires relatively little know-how in image processing.
Before being used for Visual Quality Control, artificial neural networks need to be systematically trained and validated. However, for this as well as for the initial selection and final monitoring of the algorithm in operation, the involvement of experts and professional software for industrial Machine Vision is recommended.