What makes smart cameras an ideal solution for machine vision and inspection?
June 7, 2013
By Ben Dawson
Smart cameras combine a digital camera, processing, software, and communications in one small package.
These cameras move vision processing into the camera and return results rather than—or in addition to—images. They are an excellent choice for a machine vision system to do tasks such as gaging or visually detecting part defects.
Smart cameras continue to become faster, smaller, and easier to integrate. Their software continues to improve and expands the smart camera’s applicability beyond manufacturing quality control.
It’s the software
Both PC-based machine vision systems and smart cameras require programming because they are general-purpose tools. If the programming takes too much time, the need and money for a machine vision project evaporates. Easy-to-use software is therefore required for rapid and successful development of a machine vision application.
Modern machine vision software allows you to program your vision application graphically—point-and-click to select a tool, such as a caliper, and then position that tool on the part to be measured. There is no programming in the usual sense. Vision operations are presented in familiar terms, rather than the terms of machine vision algorithms. For example, the caliper tool is presented as if it was a mechanical gage so there is no need to understand sub-pixel edge detection algorithms.
New smart cameras are designed to be servers on an Ethernet network and have no way to display images themselves. Instead, images are sent over the Ethernet to a client PC. Here a technician can set up and program the smart camera from the client PC and then log off of the smart camera and let it run independently. Results are reported over Ethernet using standard industrial protocols or via digital input/output lines from the camera. One client PC can manage many smart cameras.
Replacing human vision with machine vision
Replacing human vision with machine vision can reduce costs and improve product quality on tasks that are fast, precise or repetitive.
Machine vision systems have three general capabilities:
• Location or search finds the position of parts and guides automated assembly
• Identification of parts, perhaps by shape or an optical code (e.g. a bar code)
• Inspection checks for specific defects, such as the wrong part dimensions
First, ask if your task can cost-effectively be done by machine vision. Tasks where you cannot control part dimensions, presentation, or lighting are difficult to automate. For example, currently it is more cost-effective to use people to pick and pack strawberries than machine vision.
Next, work with your machine vision vendor and/or systems integrator to define what the machine vision system needs to do. For example, specify acceptable part variation and how part defects appear in an image. Experimentation with machine vision system components including lighting, part position, optics, and vision algorithms, is necessary for this definition.
Is a smart camera appropriate?
Then ask what kind of machine vision system is appropriate. Of course the system must have sufficient resolution, repeatability and precision, must be fast enough, and be within budget. Less obvious is the need for an easy-to-use interface to the machine vision algorithms that compute measurements from digital images of the parts.
Many machine vision tasks can be done with a “smart camera.” A smart camera packages lens, camera, processing (CPUs and software), communications, and sometimes lighting into a small, rugged package.
A conveyer presents cards with pills in the camera’s field of view. Diffuse lighting allows the camera to “see” through the plastic packaging so it can locate each pill, identify incorrect pills, and inspect pills for cracks or other damage. The BOA communicates with a PLC (programmable logic controller) to reject assemblies with incorrect or damaged pills. The camera is a “server” on an Ethernet, and so is programmed by another “client” machine on the same network. Results are also displayed on the client machine.
When to use a smart camera:
• When size, cost, and /or power are issues
• Moderate pieces per minute
• Moderate computation
• When easy to use software is included
Smart camera tech tips:
• Smart cameras are appropriate for the many machine vision tasks that require two- or three-dimensional imaging and have moderate computation and speed requirements.
• Easy-to-use software is required for rapid and successful development of a machine vision application.
• The convergence of smart mobile devices and machine vision allows small, fast and inexpensive machine vision systems that are perfect for many machine vision tasks.
Ben Dawson is director of strategic development for Teledyne DALSA’s Industrial Products Division, overseeing algorithm and product development. A former research scientist at the Massachusetts Institute of Technology (MIT), he has authored more than 60 scientific and technical papers on the subject of human and machine vision. Dawson earned his MSEE and Ph.D. degrees from Stanford University. For more information, visit www.teledynedalsa.com.