When is a smart camera not a smart decision?

Posted by Gretchen Alper on Thu, Jun 23, 2011

There’s been a lot of ‘buzz’ in the machine vision industry about smart cameras.  Why wouldn’t anyone be interested in something that is ‘smart’?  As with smart phones, there is no standard definition for this category of camera. But it is generally agreed that a smart camera does something more than just capture an image, using a sort of embedded intelligence to perform relatively simple and predictable functions, such as bar code reading or label alignment.  Smart cameras usefulness is best suited for end users that need to make simple go/no go decisions.  They are very good at performing repetitive tasks in an automated and consistent way, and are typically a single-function, self-contained system.  Smart cameras combine visual acquisition functionality with limited and focused data processing capability, all of which is contained within the camera itself (i.e. with no links to other data, components of the inspection systems, or operational systems in the enterprise).

Given those characteristics, it should be clear that smart cameras are unlikely to be sufficient for applications that are either more complex or more variable in nature.  So if a machine vision solution needs to be scalable, adaptable to changes, customized for specific needs, and /or integrated within a larger inspection system, then a smart camera may not be the best option.  In addition, performance and integration limitations present challenges for more advanced machine vision applications, such as semiconductor wafer inspection, for example.  For these types of instruments, the decision making/measurement software is intricate and the key differentiating factor. 

For OEMs, a camera that can do certain level of processing such as frame averaging or region of interest may offer more value to the system as it allows for more efficient analysis.  But, this does not mean the camera does the critical processing of the system, or performs ‘the algorithm’.  These ‘intelligent’ cameras can increase throughput, reduce wasted data, and decrease system costs by feeding pre-processed image data into the overall measurement algorithm.  Intelligent or high-performance cameras are likely to be a better option than smart cameras for use in sophisticated instruments used in demanding manufacturing where the measurement method distinguishes the system. 

For other middle range systems, smart cameras will still not be enough and more sophisticated software will be required separate from the camera.  While the price points for an off-the-shelf smart camera might at first appear appealing, it is important to consider the long term ROI of your machine vision system. If you anticipate frequent changes, updates, expansions, and require high speed operation and highly accurate inspection of complex objects and processes, then the cost of replacing smart camera on a regular basis is far greater than the initial investment in a robust, modular and integrated machine visions system. On top of that, the payback from the intelligence that can be derived from such a solution justifies the cost in very short order. 

Our friends at Boulder Imaging have produced a very comprehensive white paper on this topic, which outlines the pros and cons of smart cameras compared to other options for complex machine visions needs. It’s good reading and can serves as a guide for better understanding your needs. You can check it out here.

So while smart cameras can indeed be smart, their value is definitely not universal across all types of machine vision applications. It’s best to understand what your needs are today – and tomorrow –  and educate yourself on the value that a high performance camera or a more purpose-built solution can bring.

Topics: Vision System Optimization

Previous blog:

Next blog: