Vision Advances Improve Optical Inspection
- Friday, 01 August 2008
Recent advances in motion control and machine vision technologies present tremendous opportunities for Automated Optical Inspection (AOI) systems. Thanks to recent developments in these fundamental building block technologies, today’s AOI systems can carry out inspections with a higher resolution and accuracy, and with much faster throughput than before.
High-speed AOI systems are now being used across a wide range of industries including semiconductor and electronics, automotive, aerospace, medical device manufacturing, microelectromechanical systems (MEMS), and precision mechanical components and assemblies. But these opportunities have come with significant challenges as well. To determine the most appropriate solution that provides the best results for a given application, it is important to find the proper trade off balance.
In the case of motion systems, significant recent improvements include higher speeds, steeper acceleration and deceleration profiles, finer positional control, and faster convergence properties. Under the right circumstances, all of these factors can be used to achieve faster and more accurate point-to-point displacements and thus increase overall throughput.
In the case of machine vision technology, some of the most significant recent developments include higher resolution cameras and improvements in LED illumination. The availability of higher resolution cameras has allowed AOI systems to increase either their resolution or their field of view.
When these higher resolution cameras are used to carry out inspections over the same field of view, the system’s resolving power is effectively increased, and features can be inspected in finer detail. And when they are applied to obtain the same overall resolution, the system’s field of view can be increased proportionally. This can be achieved by using lower magnification optics, which have the desirable property of an increased depth-of-field. This is an example of technological advances that reinforce each other, as an increased depth-of-field both reduces the need to carry out system re-focusing and also increases the speed of auto-focus operations.
LED illumination modules have also progressed tremendously in recent years. They are now available in increased intensity and in varying wavelengths. LEDs are now the preferred light source in most AOI applications as they are very stable over time, have a long life, can be turned on and off very quickly and controlled accurately.
There are many more instances where improvements in one area increase the requirements in another. For example, faster motion systems with steeper profiles produce significantly more system vibrations, which lead to “motion blur” in captured images, reducing the accuracy and robustness of image analysis operations. The “brute force” approach to overcome this is to increase the system settling time, but this effectively negates the faster motion system’s benefits.
A better approach, beyond mechanical system dampening, is to increase the camera’s shutter speed. But this in turn requires a much higher light intensity to produce a still-usable image over a shorter exposure time. In spite of the recent improvements in LED illumination technology, it is still possible to reach conditions where the system is light-starved. There are limits to how fast a shutter speed can be used, just as there are limits to the capabilities of all of the individual components that make up an AOI system, in spite of all of the dazzling advances.
To keep up with the increased throughput of the faster motion systems and to deal with the much larger stream of data coming from the higher resolution cameras, significant increases in processing power are required. Fortunately, computers and workstations have themselves been progressing steadily. But many recent performance improvements in computers and operating systems have come on the fronts of multiprocessor, multi-core and hyper-threading enhancements. All these approaches do indeed yield more processing power, but only if the software is tailored accordingly. Standard serialized algorithms running on these parallel environments only get a fraction of the benefit.
To fully profit from recent performance improvements in computers and operating systems, classic algorithms need to be parallelized, optimized parallel algorithms need to be developed, and code and operations need to be made concurrent. Without this, an AOI system will not be able to reap the full benefit of all of the recent advances in motion control and machine vision technologies.