Computer vision in industrial companies

In industry, more and more companies are relying on computer vision systems that can recognize and process objects in still and moving images. They use them, for example, to automatically detect visible damage to machines or infrastructures for maintenance purposes, to inspect workpieces produced for defects for quality control purposes, or to determine parts and equipment for reordering for their inventory management. A critical component in the application of computer vision is the recording technology, because its success depends to a large extent on the image quality. Different cameras and laser systems are available for capturing images, each with their specific strengths and weaknesses. Which solution is best depends on the specific application. 

Swiss data science specialist LeanBI explains which factors industrial companies need to consider when making their selection:

What are the lighting conditions? 
Cameras require very good illumination, especially for moving parts, so that the images do not become blurred. Laser systems, on the other hand, can work reliably and deliver accurate data even in unfavorable lighting conditions.

Is there any contamination? 
In industrial plants, contamination often has a negative effect on the image quality of cameras. Cleaning systems using compressed air or water can eliminate the problem, but they make the overall system more expensive and more prone to errors.

How accurate do the recordings need to be?
Standard cameras with 4K resolution are inexpensive, but often not enough in industry. Submillimeter imaging requires expensive cameras with resolutions of 10K and more. Laser systems that are inherently high-resolution can then be the better choice. 

Is three-dimensionality necessary?
Cameras capture 2D images, laser systems can generate 3D images. Although 3D models can be reconstructed retrospectively from 2D images using photogrammetry, this is a relatively complex process that often requires manual intervention.

Is a combination with other data useful?
Often, the combination of image data with further data, for example from acoustic and vibration sensors or current measurements, can offer added value. This added value also depends to a large extent on the imaging technology used. 

"Industrial companies should check and weigh the possibilities and limitations of the various techniques at an early stage," explains Marc Tesch, owner and CEO of LeanBI. "A supposedly simple computer vision application with a camera can suddenly become complex and expensive if it turns out that good illumination and a cleaning system are required. In such cases, a laser system may be a better solution from the start."

Source and further information: www.leanbi.ch​