Intelligent Imaging Yields Smarter Machine Vision Solutions

The processor in Sony’s latest smart camera runs at 1 GHz and has 512 Mbits of memory, providing more than three times the processing power of the company’s initial 2005 models.

In recent years, the smart camera has revolutionised the machine vision sector, so much so that the product group is still the fastest growing segment within today’s market. 

Removing the need for an external PC to process video data meant an increased robustness necessary for industrial applications. Tasks could more easily be implemented, and locally pre-processing images delivered a reduction in the volume of data being moved around a distributed system by several orders of magnitude.

Previously the smart camera had been a fixed-function module and only commercially viable for mass-market applications. Recent generations, however, have helped to redefine the concept, using an open platform and allowing custom developments to be created on the fully integrated, robust, and small camera system. Essentially, this is an embedded PC with a built-in camera capable of bespoke applications and viable in even the most niche of business sectors.

The newest smart cameras have pushed this concept further and deliver an essential competitive advantage for the organisations using them. In particular, an increased processing speed, easy to use GUIs, colour images, and faster connection speeds have led to vast improvements over the original models.

An Evolution In Progress

There is large disparity between the high-end and low-end smart camera modules available to the market today. This is caused, in part, by the technology sector’s fast market growth as innovation is happening rapidly and a host of smart camera products have emerged in recent times.

Colour modules and an increased level of flexibility have been introduced to the smart cameras. For example, C and CS lens mounting cater to different applications. Also, users can easily control and stipulate the function of an increased number of inputs and outputs to best meet their requirements.

Plus, of course, the most important evolution is the increase in processing power that’s now available. The processor in Sony’s latest model runs at 1 GHz and has 512 Mbits of memory, providing more than three times the processing power of our initial 2005 models (see the figure). And this is set to increase further.

In addition, in 2008 we coupled the CPU with an open FPGA, providing pre-processing for real-time raw image data without any CPU load. This enables CPU processing abilities to be redirected toward higher-level operations, delivering faster video analysis and reducing the camera’s power consumption.

Image sensor output is now 30 1280-by-960 SXGA images per second. In contrast, the early C.2005 generations of smart cameras could only manage five to eight images per second. These evolutions have opened up the number of industrial applications that can benefit from smart camera technologies.

Quality Control

Being able to reliably and automatically detect manufactured products that fall outside a tolerated quality level, before they leave the factory, is a key benefit that smart cameras bring to the industrial setting.

One example comes from a Bavarian glassware plant. The manufacturing line creates beverage bottles, and consistently high quality is essential. On the assembly line, several pressing moulds work side by side and a smart camera vision system is used to read the unique codes printed onto each bottle, linking any defective bottles back to the mould that created them so repairs can be made.

Another example comes from one of Europe’s leading liquid propane gas (LPG) companies. Based in Istanbul, the company’s storage capacity approximates 170,000 cubic metres and sells LPG in cylinders that must be exceptionally robust for obvious safety reasons to the public. To automatically monitor for any potential structural integrity issues, the company has implemented a smart camera quality control system that runs a simple algorithm to look for rust and dents.

Barcode Analysis

Smart cameras are enabling the next generation of barcodes. These two-dimensional matrix codes have become prevalent in manufacturing applications, and the new fixed-position smart camera-based scanners are rising to the challenge.

Indeed, the car manufacturer General Motors has already indicated that they would like to limit most of their future investments in barcode readers on their manufacturing lines to camera-based systems, rather than the traditional laser scanner, because of their capability to read both linear and 2D codes.

This uptake is by no means limited to GM. The analyst house Venture Development Corporation (VDC) has stated that installations of camera-based readers are growing by 20% in the Americas, compared with a growth rate of 7% for fixed-mount laser scanners.

Using 3D Measurements

Running algorithms to identify the size, shape, and colour of an object enables industrial components to be easily sorted, ensuring the right part is used for the right product. Furthermore, the smart camera also can be linked to a second camera module to deliver 3D vision capabilities and ensure selections are more accurate.

A Swedish wheel manufacturer has adopted a smart camera/Gigabit Ethernet  module-based system to perform three-dimensional measurement analysis. The company designs a large range of components for varied applications including military vehicles, roller coasters, and supermarket trolleys, and several products differ by very small variations in depth, diameter, or tread type. Using a calibrated smart camera-based system has reduced the overall system cost and size, eliminating the need for an additional PC.

Next-Generation Applications

As the computing power increases, the biggest limiting factor in developing new applications will be the imagination. Indeed, one such sector that could, in the near future, benefit from the new smart camera functionality is the food packing industry, enabling it to automatically detect food contamination and spoilage.

Algorithms for interpreting food behaviour and the resulting light scattering currently demand vast computing power. But the processing power and modelling capabilities are improving rapidly, meaning these models are becoming more accurate.

Furthermore, smart camera systems are typically fanless. As a result, they can maintain the sterility needed for a food production environment.

Future Innovation

The technology is still evolving. Looking forward, an increased level of integration is on the cards for smart cameras, with an inbuilt zoom lens and flash functionality among the most likely candidates.

Furthermore, the small size, open platform, and ever-increasing processing speeds are vital for the rapid uptake of smart cameras. But herein lies a paradox as faster CPUs, 2 GHz and above, require fans and therefore additional space to prevent overheating. This would limit the device’s use in several areas such as food packaging (sterility) or measurement (vibration).

New processors, such as the Intel Atom and AMD’s Athlon, are being developed to run more efficiently without losing processing power. Consequently, they generate less heat and are a temporary solution to the problem. The inclusion of the FPGA also addresses this issue, enabling more functionality with less CPU power. Heat dissipation in a fanless system needs to be kept in mind, though, as processing capabilities evolve.