During the last 15 years, machine-vision technology has matured substantially and become a very important tool for manufacturing automation. Today, machine vision is used routinely in many industries as a noncontact means of inspecting assemblies, measuring dimensions, identifying parts, or guiding robots or other machines used in assembly operations.
Machine vision is much more than an after-the-fact optional inspection step used to reject bad parts. It is particularly successful in applications where it guides a machine or closes a control loop.
Some key trends have driven the proliferation of vision systems:
Cost.
Performance.
Algorithmic robustness.
Ease of use.
Faster and Faster Hardware
Since its inception, the machine-vision industry has been characterized by the continuous introduction of faster hardware and better price/performance ratios. This has followed the same trends of the semiconductor industry that has seen desktop PCs move from 8-MHz 8086 CPUs to 500-MHz Pentium IIIs today and projections of 64-bit processors running at clock rates in the gigahertz range tomorrow.
Faster vision-processing hardware has been a key to both higher parts-per-minute throughput and greater robustness. The increased throughput and robustness have increased the number of complex inspections that machine vision can address.
In the dawn of machine vision, binary processing of low-resolution black-and-white images was all that the hardware would allow. Today, more sophisticated image processing and analysis performed on high-resolution gray-scale images are commonplace. Computationally intensive image-preprocessing operations like mathematical morphology now are widely used on hardware that allows many processing passes through an image during a single frame time.
While the standard broadcast TV frame rate of 30 Hz has been considered as real-time, the introduction of digital machine-vision cameras running at higher rates has made clear the need for faster-than-conventional-video real-time processing. Again, the latest generation of vision-processing hardware supports image acquisition from nonstandard cameras and is fast enough to keep up with the processing of more data contained in the higher-resolution images or conventional resolution images that are coming in much faster.
The advances in semiconductor technology of the last decade have enabled custom vision processing systems to continuously shrink in size from cabinets full of boards, to single boards, down to custom silicon chips. This custom vision processing hardware also continues to be key in delivering robust vision functionality in less complex and lower-cost configurations closer to the manufacturing process.
PC-Based Vision
At the same time, the ever increasing speed of standard PCs has advanced the trend toward greater acceptance of PC-based vision systems. Vision systems based on standard microcomputers have come a long way since almost 10 years ago when they were introduced by Automatix, a predecessor of RVSI Acuity.
With PC-based vision systems, both the supplier and the user can leverage a variety of third-party hardware and software. Also, there is wide acceptance of the PC in desktop and manufacturing-floor applications. Today, PCs running under the Microsoft Windows NT O/S are fast becoming a dominant platform for delivering factory-floor monitoring and control applications.
The wide availability of low-cost frame grabbers and image-processing software packages makes it possible to build vision applications using third-party components. Although feasible, this approach has its caveats, which include limitations of conventional multimedia frame grabbers, nondeterministic performance, and sometimes excessive development, installation, and support costs.
Plug-in Vision Engines
A recent development is the single-board PC plug-in vision engines and powerful component-based software environments for developing and deploying vision applications. These vision engines incorporate the functionality of a complete vision system in a single PCI board. This completely off-loads vision-related processing from the host PC. The CPU then can be dedicated to other tasks such as production monitoring, control, or user interfacing. A vision-engine configuration also off-loads the host PCI bus since all high-bandwidth image-capture operations are internal to the board.
Since the vision board fits inside a host PC used for other purposes, this is truly a zero-footprint solution—a key consideration in many OEM or clean-room applications. By plugging multiple vision boards into one PC, each can be dedicated to different inspection tasks.
In addition to on-board, high-performance CPUs and custom vision-processing hardware, next-generation vision engines typically run under a real-time multitasking operating system that allows deterministic performance in all image acquisition, vision processing, and I/O operations. This is a distinct advantage over conventional PC-based systems that run under a non-real-time Windows operating system.
In contrast to multimedia frame grabbers, these engine boards support a variety of machine-vision input devices, ranging from conventional analog video to nonstandard digital cameras. By being designed for machine-vision applications, these products include options that are not always available on conventional multimedia frame grabbers, such as strobing or channel switching between successive frames. On-board display capabilities offer the option of showing images and graphics in a dedicated display or in a picture-in-picture fashion on the host PC/Windows display.
Finally, on-board network connectivity simplifies deployment in manufacturing floor networks/intranets. It also supports innovative remote monitoring/diagnostics options that can be performed over the web.
Ease of Use and Deployment
Multiple levels of users may be involved with the development, deployment, and day-to-day support of vision applications. These users range from factory-floor operators and engineers who would typically set up, install, or modify vision applications to system integrators or OEMs who would create custom vision applications or even develop new vision tools.
Most state-of-the-art vision systems today offer built-in graphical operator interfaces and comprehensive runtime/monitoring environments that allow them to select jobs, start and stop inspections, adjust inspection parameters, see detailed or summary reports and statistics, and capture and log failed part images or other data. At one level down, manufacturing engineers who configure vision applications must do so in an intuitive environment, using high-level tools as opposed to low-level image processing and analysis operations.
An approach that describes a vision application program in the form of a sequence of steps and uses high-level, application-oriented tools now is widely accepted. Implementing such tools on higher performance platforms has, over the last few years, allowed the tools to become more robust and intelligent, which limits the need for vision expertise required by the user. This trend is certain to continue in the future.
System integrators or OEMs must be able to quickly develop custom interfaces and sometimes even add to the system’s low-level functionality by working in industry-standard software-development environments. One recent development in this area, which holds the promise of much faster application deployment for system integrators and faster time to market for OEMs, is component-based software. Software components encapsulate the core vision system functionality.
In a Windows environment, these components are developed as ActiveX Controls (Figure 1). An example of a vision application deployment component would be one that encapsulates all the functionality related to training (Figure 2). Another deployment component would encapsulate all the functionality related to loading a job and starting, stopping, and monitoring an inspection. You now can create or customize vision applications by dropping such basic building blocks inside a custom GUI developed using Visual Basic or Visual C++ without knowing much about the internals of the vision system.
The Future Is Now
Advancements in hardware and software most certainly will continue and intensify in the future. Faster hardware, more intelligent tools, and better application software development/deployment environments will all promote a broader and deeper proliferation of machine vision in manufacturing.
However, through the many recent advances along all fronts of price/performance, robustness, and ease of use, we believe that vision technology now has reached a point very close to what the industry and marketplace promised a few years ago.
At the same time, the last 15 or 20 years has educated manufacturing users on what vision should and should not be used for, and these application boundaries are continuing to move outward. At this point, machine vision is not a research curiosity but rather a mature tool for manufacturing automation.
About the Author
For the past 15 years, John Agapakis has held a variety of engineering and technical management positions in machine vision and vision-guided robotics technology and product development at RVSI Acuity CiMatrix and Acuity Imaging and Automatix, its predecessor companies, Currently, he is vice president of R&D at the company. His educational background includes a Ph.D. in vision-guided robotics from MIT. RVSI Acuity CiMatrix, 5 Shawmut Rd., Canton, MA 02021, (781) 821-0830.
Copyright 1999 Nelson Publishing Inc.
May 1999