Machine vision is a proven process control tool for a variety of industrial automation applications. Traditionally, this technology integrates commercial off-the-shelf (COTS) imaging sensors, lighting modules and processors to guide, inspect or identify parts as they move along production lines. Compared to human operators, machine vision systems are fast, accurate and repeatable—improving product quality, lowering scrap rates and increasing productivity in fast-paced manufacturing environments.
While the operational benefits are clear, high-speed digital imaging expands the benefits, capabilities and use cases for machine vision even further. Thanks to their high resolution, fast frame rates and streaming capabilities, high-speed cameras enable machine vision in challenging applications that require real-time analysis or long record times, such as semiconductor manufacturing, space shuttle launches, railway inspection and more. For these reasons, high-speed machine vision systems capture what traditional vision systems cannot, shedding light on processes too small or too fast to see with the human eye.
Key Features of High-Speed Machine Vision Cameras
High-speed machine vision cameras differ from traditional machine vision cameras in several ways. In place of COTS or charge-coupled device (CCD) sensors, high-speed cameras integrate custom-designed complementary metal-oxide semiconductor (CMOS) sensors. When designed specifically for high-speed applications, CMOS sensors deliver unmatched speed and sensitivity, leading to more detailed inspections and higher yields in machine vision applications like defect detection, part measurement, and more. High-speed machine vision cameras that incorporate CMOS sensors with multiple megapixels achieve exceptional image quality even at challenging frame rates.
It is not enough for machine vision cameras to simply go faster; they must be designed to handle it. In addition to fast frame rates exceeding 67,000 fps, the following features are important to keep in mind when selecting high-speed cameras for machine vision applications:
Resolution: Phantom machine vision cameras incorporate CMOS sensors up to 9 megapixels. These multi-megapixel sensors—coupled with the cameras’ small pixel sizes—yield more detailed images at fast frame rates.
Light sensitivity: Generally, the smaller the pixel size, the greater the imaging detail—which is especially important in applications requiring a microscope. CMOS sensors feature pixel sizes as small as 5.6 micrometers, resulting in high native ISO. As a result, these cameras achieve excellent image quality despite the low exposure times necessary for high-speed machine vision applications.
Exposure time: Phantom cameras have exposure times as low as 1 microsecond. This capability—along with high light sensitivity and small pixel sizes—sufficiently freezes high-speed motion while eliminating motion blur.
Dynamic range: Dynamic range comes into play when an image has a lot of shades, or when a subject is almost the same color as its background. The higher a camera’s dynamic range, the more shading definition the sensor can detect. Phantom machine vision cameras have a dynamic range between 54.8 and 59.7 decibels, making them suitable for darker applications like semiconductor inspection.
In addition to highly customized sensors, high-speed machine vision cameras utilize copper CoaXPress (CXP) cable technology, enabling them to transfer vast amounts of data to compatible, industry-standard backend frame grabbers in real time. This ability to stream data instantaneously avoids the time-consuming process of saving data to the camera’s limited RAM before downloading it to a computer. When coupled with off-the-shelf DVR units, these streaming cameras also support longer recording applications in aerospace, such as rocket dynamics, airplane dynamics and ballistics, just to name a few.
The CXP6 protocol is currently the fastest standard data transfer method. Each copper cable achieves data transfer rates of 6.25 gigabits per second from the camera to the backend receiver machine. The more recent CXP12 standard doubles this rate, making both CXP6 and CXP12 ideal for cameras requiring high throughput.
While most machine vision cameras provide up to 2-gigapixels per second of data throughput, the world’s fastest streaming cameras achieve direct data transfer speeds up to 9- gigapixels per second. These cameras divide and transmit images by rows, then stitch each image back together using a simple algorithm— making higher frame rates and resolutions possible. By relying on GenICam, a generic programming interface, these cameras also facilitate configuration and integration into existing systems.
Real-Time vs Long-Record Configurations
The backend configuration in highspeed machine vision systems depends on a number of variables, including the required frame rate, resolution and record time. For real-time analysis, users can utilize up to 16 standard CXP6 channels on the streaming camera. While CXP6 cables enable communication up to 68 meters, fiber-optic connectors achieve longer distances up to 200 kilometers. Users can also utilize the camera’s general-purpose input/output (GPIO) for fast, flexible signaling and synchronization.
Other hardware and software components include:
App-specific frame grabber and CXP6 frame grabber card;
Software, including the frame grabber API, Matlab®, LABVIEW or any postprocessing vision tools;
Image processing hardware, typically a GPU microprocessor or FPGA;
A computer with PCIe (Peripheral Component Interconnect Express) Gen3 slots.
While most high-speed applications take place in hundreds of milliseconds, machine vision applications may require longer recording times—several minutes to half an hour, for example—to accommodate events like space shuttle launches. To overcome the storage challenges associated with such a large amount of data, high-speed machine vision cameras can stream data directly to a DVR unit, which features multiple terabytes of space. This plug-and-play setup lets users easily store the incoming data for analysis at a later time.
Expanding Traditional Machine Vision Applications
Thanks to high-speed machine vision cameras, manufacturers can achieve imaging with higher resolution and recording accuracy. These features increase line speeds and production volumes, decrease bottlenecks and reduce costs per unit. At the same time, these streaming cameras can capture nanometer-scale targets that are otherwise difficult to see—let alone analyze—using traditional machine vision cameras, transforming machine vision from a process control tool to a diagnostic tool. As a result, machine vision is gaining traction in industries like the life sciences, semiconductor manufacturing, pharmaceuticals and more.
Some of the latest application areas for high-speed machine vision include:
Semiconductor inspection: High-speed machine vision cameras are playing an increasingly important role in semiconductor manufacturing, an industry driven by throughput. Specifically, they quickly identify and flag part defects as soon as they arise, lowering defect-related costs and downtime, improving throughput, and keeping inspection times to a minimum.
High-quality machine vision cameras strike the necessary balance between light sensitivity, signal-to-noise ratio and fast frame rates required by semiconductor applications, which typically involve submicron size scales. At full 4,096 × 2,304 resolution, the Phantom S990 streaming camera, for example, features 6.75-μm pixel sizes, 9.6e- noise and 938-fps recording speeds, generating high-quality images that allow the imaging software to detect subtle variations between the light and dark areas indicating a defect.
High-speed spectrometers: Used in many food and beverage, pharmaceutical, and agricultural applications, high-speed spectrometers diffract white light into various wavelengths to create an absorption spectrum, making it possible to detect the presence of certain materials. One emerging application for this process involves verifying the chemical makeup of pharmaceutical tablets using high-speed machine vision cameras. The cameras record tablets as they move along a conveyor. Then, based on the diffracted, absorbed and transmitted light that reaches the lens, the camera can flag defective tablets, all while keeping production lines moving.
Thanks to their fast recording speeds and light sensitivity, high-speed streaming cameras are ideal for this emerging application. In addition to providing a non-contact inspection method, these cameras can easily inspect dozens of tablets within the same field of view simultaneously, improving throughput in critical pharmaceutical operations.
Streaming Camera Features
In machine vision streaming applications, image data flows directly to a frame grabber and PC or long-record DVR via CXP cable technology. Users can immediately access this data either for a real-time application or long record and are limited only by the amount of storage in the PC or DVR.
Other features of high-speed streaming cameras include:
Configurable bit depth—8-/12-bit, 8-/10-bit
Powered by CXP6 technology for some streamers that consume less than 27 volts
GenICam compliance for easy integration
Compatibility with PCIe3 CXP6 frame grabbers
GPIO provides common and advanced signaling features
Scalable data transfer for reduced data needs
Railway inspection: High-speed machine vision cameras have the potential to change how railway systems are inspected. Unlike traditional machine vision cameras, which lose valuable time during the year to rain, snow or dust storms, streaming cameras provide the high frame rate, resolution and light sensitivity necessary to see through difficult weather conditions. Unlike most cameras, they also handle white light without the use of additional lens filters.
Other emerging application areas for streaming cameras include the life sciences, such as cell diagnostics; vial identification and hemolysis; ballistics testing; laser welding; and 3D printing.
This article was written by Uma Gobena, Vision Application Engineer, Vision Research (Wayne, NJ). For more information, visit here .