Breaking down barriers to machine vision (continued)
A continuation of our interview with Laval Tremblay, VP of Engineering, Matrox Imaging, which appeared in the November 2007 Viewpoint column.
By Larry Maloney, Contributing Editor -- Test & Measurement World, 11/1/2007
|
Q: Which communications protocols are dominating in machine-vision applications?
A: There’s no question that analog has lost ground to digital, but analog is still a field-proven, robust, and inexpensive technology, so sales are still strong in the analog channel. As for digital interfaces, Camera Link and 1394 are the most widely used today, though Camera Link offers the most bandwidth. Not only is Camera Link the king of data throughput, but it’s a superior deterministic protocol. Engineers turn to 1394 as a cost-effective solution for applications that don’t have the bandwidth or determinism requirements of a Camera Link application. However, GigE Vision is becoming a serious challenger to 1394.
As for how to choose the most appropriate interface for an application, bandwidth is a limiting factor for most protocols, so you must first define the resolution, frame rate, and pixel format to determine the bandwidth. Then you can ask the camera vendors if what you need exists for a particular interface standard. You will also have to consider other factors, such as determinism, data reliability, and cable length.
Q: What are the target applications for your new Vio line of frame grabbers?
A: The Matrox Vio was designed primarily for high-definition medical imaging and high-definition video surveillance. Its single-platform capture card supports analog HD/SD input and output. Some of its key features include low-latency video output (synchronized to the video input), 128-Mbyte onboard memory, support for 20-bit video, automatic detection of the video source, and graphics overlay composition.
Q: What other new product introductions reflect where Matrox is headed with its machine-vision technology?
A: The media in this industry keep predicting the death of the frame grabber, but customers are continuing to buy frame grabbers in record numbers. But even so, we’re developing products that solve current technical issues. The Matrox Solios GigE frame grabber is a good example. Not only are we supporting a digital interface with great potential, but we’re including unique features.
Sure, developers can use a standard network interface card for a camera, but managing thousands of packets and reconstructing images can tax your system’s resources. We’ve designed the Matrox Solios GigE to handle packet management and image reconstruction, so that the PC can take care of the image processing, which is the important part of the application. But more than that, the Matrox Solios GigE offers other machine-vision-specific features, such as color space converters, look-up tables, processing FPGAs, general-purpose I/O and long life cycles. These are features that standard consumer-based interface solutions just don’t offer.
Q: What progress do you see in the use of color in vision applications?
A: There is a lot of interest in color, and the food and beverage industries are leading the way in terms of usage. Most industrial vision applications, however, still rely on monochrome images. Color is relatively new in terms of commercially available algorithms, but we’re beginning to see more and more color image-processing tools.
Q: What are the chief strategies that you use to address lingering customer doubts about machine vision?
A: We have always tried to lower the cost of components. But price isn’t everything. It’s important for us to develop software tools that OEMs and end users will find easier to use. Generally, that means more configuration, simplifying parameters, and less programming. Our Design Assistant’s flow-chart-based development environment simplifies the whole process. That’s the kind of innovation that can bring machine vision to a wider audience.
Read the first part of this interview.
