The low-voltage, high-performance OV7962 provides the entire functionality of a single-chip digital VGA sensor and has extremely low light sensitivity. Although its VGA resolution is on the low side for traditional machine vision, the chip incorporates advanced image processing functions, such as automatic exposure/gain control, lens correction, defective pixel correction, white balance, and horizontal and vertical windowing capability.
As a reference module, the SmartVue gives developers the ability to produce a proof-of-concept design for testing algorithms and applications early in the product development cycle. It comes with a comprehensive software suite: the Image Cognition Processor Software Development Kit, the OpenCV-based Imaging Tool Kit, various application tool kits, and code development tools.
The camera module is expected to enable the development of several different types of embedded vision applications that execute real-time image processing and scene analysis. These include automotive driver-assistance systems, control of consumer appliances with gesture recognition, using face/people detection and tracking to measure audiences, smart video monitoring, asset monitoring, sensing occupants in "smart" buildings, military uses for body-worn cameras, and robotics.
Although traditional industrial machine vision inspection applications aren't mentioned in the company's list of target applications, there's a parallel trend that may well take advantage of this tiny embedded smart camera. That's the growing use of smaller and smaller portable inspection systems that can be moved around the factory, for uses such as incoming and outgoing inspection.
I said earlier that embedded vision is a big, yet mostly unknown, trend in traditional machine vision. The Embedded Vision Alliance, founded in May 2011, was created specifically to address this trend, which is the result of converging technologies and of shifts in image sensor price/performance thresholds. As Jeff Bier, president of founding member company BDTI, said in an article I wrote for Test & Measurement World, embedded computer vision remained in low-volume applications like industrial machine vision until recently because it was based on historically expensive, low-volume, specialized technologies. But around 2005, image sensor performance went up as prices came down, much of this fueled by the success of cameras in mobile phones. And then came the typical semiconductor/electronics technology curve: Volumes go up, prices come down, volumes go up again.
I've been writing about technology long enough to not get too excited about new product or technology announcements. But this one is different. I think embedded vision is about to become a very big deal.