Focusing on smart cameras

Smart cameras combine high-performance image sensors with a built-in processor, alleviating the need for processing on an external PC.

A decade of rapid development in technology means that devices in every industry are smaller and more powerful than their predecessors by an order of magnitude. Take a look at your mobile phone for example - today it has more processing power than a computer of the early 1990s, and typically combines a camera, MP3 player and a Web browser all into a single handheld device.

In the machine vision sector, these advancements have fuelled the development of embedded image processing devices capable of running algorithms closer to the image sensor of a camera. A smart camera is an industrial vision system that combines a high-performance image sensor with an inbuilt processor, enabling measurements to be taken directly on the camera without the need for processing on an external PC. A smart camera returns the result of an inspection, not merely the image itself (although it can do that as well) and it can make other intelligent decisions to drive actions in an industrial environment. They are sophisticated enough to communicate with a variety of other distributed devices over a network, and typically include lighting control and industrial I/O as part of the standard feature set of the camera.

Integrating each of the components of a vision system onto the camera shrinks the size and cost of a vision system dramatically. The smart devices are easy to program, deploy and maintain in a production environment and can be integrated with other industrial devices such as PLCs, programmable automation controllers (PACs) and human machine interfaces (HMIs). The cameras are designed to have a compact, rugged form factor, which makes them ideal for mounting in harsh industrial environments and enables them to remain reliable under extreme conditions. Both the small size and the durability of these devices simplify the mechanical details of integrating a smart camera with other machinery on the production floor. With all of these benefits, smart cameras have become increasingly popular and are now available from a number of vendors as commercial-off-the-shelf (COTS) products.

Components of a Smart Camera

Compacting all of the features of a full-blown vision system into an all-in-one rugged camera packs a lot of power into a small, compact device. Smart cameras incorporate an image sensor, lens, processor, lighting control, industrial I/O and ethernet communication all in a single device, which also runs the vision algorithms for your inspection system. The exact features of each component will vary with the specific smart camera and in many cases offer added performance bonuses over a traditional vision system.

Image Sensor and Lens

At the core of any vision system you will find an image sensor to acquire images for analysis and processing, and smart cameras have an image sensor integrated into the heart of the electrical architecture of the camera. The sensor is the same as you would find in a typical industrial camera, a high-resolution CCD or CMOS sensor which produces VGA or monochrome images. The maximum resolution of smart cameras will vary depending on the make and model of the sensor used, with resolutions starting around 640x480, at 8 bits per pixel. The frame rate of the camera is also a factor of the sensor used, ranging from 30 to hundreds of frames per second.

The lens choices for a smart camera are no different than that of standard cameras. Many of them have a rugged c-mount lens connector directly in front of the image sensor to allow you to pick a lens with the proper aperture and focal length for your application. The combination of image sensor and lens field-of-view and focal length is chosen to produce sharp images with sufficient resolution to increase the accuracy and effectiveness of vision algorithms running on the embedded processor.

Embedded Processor

Sitting next to the image sensor on a smart camera, you will find a high-performance processor that runs image processing and analysis algorithms directly onboard the camera. Adding a processor to the camera has the benefit of simplifying the required computational elements of your vision system. A smart camera uses the onboard processor to make decisions in the camera and output the correct measurement data to the rest of the system, as opposed to a PC-based vision system which would require the transferral of the raw images to another computer for analysis. The processor on the camera increases the speed at which the computation is made, as well as reducing the overall system cost and complexity.

A typical example of the embedded processors found on a smart camera is the Freescale PowerPC. This processor is capable of running the entire suite of image analysis and processing routines, that one would expect to see from any vision system, directly on the camera. Using a real-time operating system, a smart camera can deterministically analyse images and deliver the results to other system components in a timely manner. You can also run other software on the processor, such as graphical configuration-based tools, making it easy to program advanced vision algorithms on the camera.

Embedding processors inside each camera in a network of distributed smart cameras gives additional processing advantages for a vision system. Instead of relying on a single external PC to do all of the number crunching, each individual smart camera can perform the required analysis at each node in the system. By returning only the results of the image inspection, you drastically reduce the bandwidth requirements between devices on the network. A processor on each individual smart camera increases the raw processing power of the overall system versus a single external processor in a PC.

The onboard processor can also be used to compute information for other devices such as PACs or HMIs. An external PC may still be needed to perform simple tasks or log results, while the smart cameras themselves are used to perform the brunt of the image processing, taking advantage of distributed computing throughout the vision system. 

Lighting Control

Automated inspection of an image with poor lighting can involve more image processing steps, reducing performance and, in the worst case, introducing errors in vision algorithms that, for example, perform basic operations like edge detection or shape detection on a unit under test (UUT). To illuminate objects for inspection, smart cameras incorporate communication for controlling external lighting sources from the camera. A common example involves trigger signals that are sent from the smart camera to a lighting controller to cause an LED to strobe as UUTs pass in front of the camera.

In some cases, a smart camera is capable of driving the lighting itself. For example the National Instruments NI 1742 Smart Camera has a feature called direct drive lighting technology, and includes an inbuilt lighting controller, lowering cost and simplifying wiring. The direct drive lighting controller can provide a constant DC current of 500mA, as well as strobed current of up to 1A, capable of supplying most current-driven LED light heads. With strobe lighting, you can increase the intensity produced by up to four times without harming the light head. Many cameras include mounting points for ring lights or other lighting fixtures on the outside of the smart camera around the lens for uniform illumination of a UUT, allowing lighting to be easily incorporated onto a camera.

Industrial I/O and Ethernet Connectivity

Smart cameras provide a variety of connections, including digital I/O and encoder inputs. Many support ethernet-based and RS232-based industrial protocols to communicate inspection results between industrial devices. Digital I/O ports allow for direct connectivity with triggers and actuators, while encoder inputs enable a smart camera to synchronise inspections with linear and rotary drive systems. Ethernet ports can be connected to an industrial network to send data between multiple smart cameras, or to integrate with other devices such as PACs for expansion I/O, or HMIs to create interactive workstations for displaying results on the production floor.

Vision Algorithm and Device Software

With all of the components integrated into a smart camera, you would think programming such a device would be more complex. That is not the case. Software is the element that really differentiates a smart camera from a traditional vision system. Smart cameras are meant to be easy to get up and running quickly for basic inspections and image acquisition.

These smart devices can run an entire suite of vision algorithms on its embedded processor directly on the camera, allowing it to perform pattern matching, optical character recognition (OCR), colour inspection, image arithmetic, edge detection, filtering, object classification, and more, directly as it acquires images in real-time. The vision algorithms packed onto a smart camera make it an all-in-one sensor able to locate, identify and inspect objects in a single device. 

To help a user get started using the vision algorithms included with a smart camera, many companies include a development environment to help them rapidly configure their machine vision application out of the box. Software such as National Instruments Vision Builder for Automated Inspection allows a user to define pass/fail criteria and classify parts through an intuitive user interface without any programming. Software tools like Vision Builder make it possible to design, prototype and deploy applications on smart cameras with a fraction of the effort required for traditional vision systems.

Smart cameras can also run advanced development environments for users who need to develop more demanding applications harnessing the full power of the feature set on the camera. The flexibility to choose between configurable inspection software and a full programming environment, gives a user all the tools they need to setup and configure a smart camera for a variety of applications.

Application Areas

The benefits of integrating each of these components together on a smart camera make it a viable option for a variety of different machine vision applications. Taking measurements to verify parts are in the proper location or the correct size is easy with machine vision algorithms today. The cameras work well in industrial environments requiring inspection of products coming off a production line. You can use a smart camera to verify the presence of a cap on a bottle, or that a label is properly aligned on a package. You can even read 1D and 2D barcodes to verify part numbers, or to automatically inventory parts used on a larger product. The cameras are also often used in sorting applications for analyzing images to determine what product is in front of the camera, and then control a motor or air nozzle to move the unit into the appropriate bin.

The combination of a high-performance processor with a high-quality image sensor on a smart camera allows engineers to easily create distributed machine vision systems that transmit inspection results instead of raw images.

Smart cameras are cheaper and easier to use than PC-based vision systems and are more reliable when deployed in the field. They save development time and deployment cost by packing all of the required components for a vision system into a single device. With all of the benefits and power packed into smart cameras today, they are sure to be a serious contender for use in your vision system for tomorrow.

Recent articles

Info Message

Our sites use cookies to support some functionality, and to collect anonymous user data.

Learn more about IET cookies and how to control them