Visionary Insights: Exploring the Impact of Computer Vision in Electronics Manufacturing (2024)

AI is already in use in some of our manufacturing solutions, such asstencil and screen printersfor the solder and adhesive materials application processandpick and place machines for theSMT assembly, i.e. the component placement process(SMT Mounter). Rapid improvements in computer vision is likely to assist in two additional areas:

1. Electronics Final Assembly

When products leave the automated production line, often additional bulky or heavy leaded odd shapedcomponents need to be inserted or sub-assemblies connected to main boards. This fiddly task is performed by people.Especially for complex boards or for production environments with frequent changeovers between a high variety of products even experienced assembly operators might need some sort of placement instruction guidance. This needs to happen in a way that the operator is not unnecessarily disturbed by confirmation actions whether a component has been placed correctly or not. Cameras with increased accuracy and latest computer vision applicationshelp to overcome these challenges.

2. Analyzing and Optimizing Operator Movements

Computer vision, backed by AI,can enable manufacturers to automatically analyse and optimise the movements of operators inside a factory and in manufacturing cells to improve efficiency. This can be for moving between various process steps, e.g. between preparation areas and machines. Or if a smaller observation area is used it helps during training or even at performing critical maintenance tasks to ensure maintenance quality.

Back at university, computer vision was the topic of my thesis and I have been following the research for personal interest. In my opinion, there is always a delay between state-of-the-art technology from research to industry,I know of cases wheree.g. technology developed in 2009 in the lab was then introduced to the market in 2015.

However, in our SMT and THT manufacturing industry the situation is a not that bit different, I believemanystate of the art technologies arenot yetused in the key processes. But for different reasons!

The basic applications are already sufficient for the current processes. As we can fix and calibrate the position of the camera it becomes easy to work with different pictures of an ideal picture by using a simple pattern matching algorithm. Computer vision is capable of working with not defined positions through SLAM (simultaneous localization and mapping).One of the core processes which enable high degree quality control within the SMT production environment, is parts recognition and 3D reconstruction. With the helpofour multi camera type 3 we can generate a3D multi-view vision. This process can be done quickly and easily as camera positions are fixed and known. Inspection machines within the industry often use at least 8 simultaneous views of the same object in order to get high multi-view visions.However, in my opinion, computer vision could be used for even more complex applications within manufacturing, because the same multi-view model reconstruction methodology could be used to generate a freehand television application. It might give us the opportunity to create videos and freely choose viewpoints which we have never been recording from. This is particularly useful for training or maintenance tasks, as it allows for the generation of virtual pictures from even the most inaccessible positions, where taking a normal picture would be impossible.