Emerging trends in machine vision
Many improvements enhance the use of machine vision, which is defined here as the automatic interpretation of images for industrial applications. These advances include faster, higher-resolution systems, easier-to-use software, and cameras with built-in processing. Moreover, the increasing use of LEDs as an illumination source reduces the cost of such systems. The Automated Imaging Association (AIA) (Ann Arbor, MI) concluded that more than 50% of illumination systems sold to the North American vision market in 2005 were based on LEDs. Overall, these advances make machine vision more affordable for tasks that previously would not have justified the cost. These trends, however, are ongoing rather than emerging.
Today, the emergence of standards is the most significant trend in machine vision. By this, I mean standards that are specific to the machine-vision industry. The first of these is the Camera Link standard for interfacing cameras to frame grabbers, which digitize analog video signals. This standard ensures easy integration of these devices, even ones from different manufacturers. Connectors, cables, data format, and control signals are all standardized. Industry-standard cables will also drive competitive pricing when manufactured in volume.
More recently, the GigEVision standard emerged. It regulates how a gigabit Ethernet connects cameras to computers, not necessarily with a frame grabber. Currently, this method is somewhat slower than Camera Link, but GigEVision provides a great advantage: the distance between the camera and computer is effectively unlimited. So in a hostile environment, only the camera needs to be protected, and the computer can be located in, say, an office. This makes GigEVision interfacing particularly beneficial in some industries, such as food processing and pharmaceuticals, where high-pressure hoses may be used to occasionally wash down the production line. The AIA hosts both the Camera Link and GigEVision standards.1
The European Machine Vision Association (EMVA) (Frankfurt/Main, Germany) hosts a standard called GenICam, which is closely allied to GigEVision.2 According to the EMVA, the goal of GenICam is to provide a generic programming interface for all kinds of cameras. The GenICam standard consists of multiple modules. The GenApi module configures the camera, and it should be the same for any interface technology—GigEVision, Camera Link, 1394 DCAM, etcetera—or features being implemented. Although the GenICam standard is not specific to using a gigabit Ethernet, any GigEVision device must provide an XML device description file that complies with the syntax of the GenApi module. GenICam's Features module includes recommended names and types for common features. The TransportLayer grabs images. Last, the DataStream module interprets additional data that might be appended to the image.
The first version of the GenICam standard contains only the GenApi module, and the others will follow. In simple terms, the camera will be able to tell the controller—located in a frame grabber or the main processor—what functions it has onboard and how these can be controlled, without the need to study the documentation.
Finally, EMVA also hosts the 1288 standard.2 This initiative defines a unified method to measure, compute, and present specification parameters for cameras and image sensors used for machine-vision applications. This standard will benefit customers and manufacturers of vision components. It will avoid misunderstandings and reduce pre- and post-support cycles. It does not attempt to standardize cameras themselves, just the way their performance is measured and described.
These machine vision-specific standards should make the technology appear more mature, especially to a first-time user who might believe that machine vision is so new that choosing one particular supplier might turn into a dead end if that supplier withdraws from the market. With these standards in place, users can mix and match cameras, frame grabbers, and processors. Moreover, various manufacturers could provide replacement parts at any time.
Don Braggins is an independent consultant in machine vision. He founded his consultancy in 1983. He is also the executive director of the UK Industrial Vision Association, which was founded 1992. In addition, he is a SPIE Fellow and chaired several SPIE European conferences in the 1990s. He also serves as the associate editor of pattern recognition and machine vision for Optical Engineering.
