TECHNOLOGY IN CONTEXT
Vision and Inspection Systems
Advanced Vision Platforms Based on AdvancedTCA Architecture
A solution for advanced machine vision applications that fulfills the demand for advanced technologies, customization and application emphasizes the reduction of operating expenses and space consumption using the ATCA architecture.
STEPHEN HUANG, ADLINK TECHNOLOGY
Page 1 of 1
Machine vision system integrators and equipment providers are constantly in search of new solutions and high-performance computers that are capable of executing complex algorithms for image processing as well as products that are cost-effective and have faster time-to-market cycles. Take the Automatic Optical Inspection (AOI) application for the Flat Panel Display (FPD) industry as an example. As the FPD industry progresses, the demand for multiple high-resolution cameras on a single system is increasing, while data throughput ability becomes critical in handling heavy data payloads.
The phenomenal growth of vision applications for large format imaging (both line scan and area scan) requires imaging, displaying, storing, real-time visualization and high-end processing. A server equipped with dual CPUs and a PCI-X add-on card for the camera interface is commonly used for AOI applications. However, when multiple cameras are required, multiple servers are needed to build an effective AOI system. These servers take up valuable space and entail a considerable amount of administration and maintenance. In addition, data exchange between servers also poses a greater challenge for system integrators and administrators.
Not only are AOI systems valuable in the FPD industry, but also across a wide array of industries that utilize visual input for data processing. Applications such as those used in navigational guidance systems for autonomous robots, positional and structural information for manipulation tasks, mapping terrain features, particularly in inhospitable environments or during natural disasters, as well as high-speed, high-volume manufacturing processes where computer-aided operations have become a necessity, are all beneficiaries of AOI technologies. AOI systems equipped with the proper human machine interface enable operators to perform real-time and accurate operational control for quality assurance, which is top priority no matter what the industry or application.
AOI systems involve acquiring images into a computer, converting the images into usable formats, adjusting the images to the desired views, and calculating the appropriate data that represent the images to perform quality inspections. When specifically applied to a manufacturing environment, an AOI system must satisfy not only the requirements for high speed, high resolution, 24-hour operation and repeatability of measurements, but also for automatic identification, tracking and quality assurance throughout the entire production process.
Some featured AOI systems with customized DSPs and FPGAs give basic solutions for translating results and locating defects. With advanced optical inspection technology, users are able to recover more of the good product (higher yield) and remove a higher percentage of defective product (quality control) than the manual sorting and defect removal methods historically used by many industries. In the wood panel industry, for example, increasing the number of decisions made by a machine vision system can also result in increased throughput, higher yield and more accurate product grading with fewer line workers, aiding a company’s bottom line by cutting costs. AOI systems can add significant value in manufacturing environments where processing is highly variable, by improving the uniformity of finished products.
However useful, without the proper computer architecture and controls, AOI systems run the risk of longer development times and effort for the integration of DSPs and FPGAs across multiple programming languages. Therefore, these solutions require higher software and hardware investments related to the acquisition and adoption of new DSP technologies.
A solution for an advanced AOI application involves a platform that delivers an optimum balance between product capacity and cost ratio in relation to processing and input/output support, easy programmability, customization and maintenance.
AVP Platform Meets High-End Requirements
Although primarily designed for next-generation telecom applications, some benefits of Advanced Telecom Computing Architecture (ATCA) are found to be of great advantage in solving the current problems of high-end machine vision systems. The open architecture saves development time and associated costs, while ATCA systems equipped with PMC card expansion enhance system flexibility. By integrating customized PMC cards and utilizing innovative Gigabit Ethernet and shelf management technologies, a machine vision system based on ATCA ably meets all performance requirements in a compact and high-density, multi-blade rackmount system.
A high-performance CPU blade with a PCI-X PMC module forms a single advanced CPU platform. A PMC module can interface with several kinds of camera interface standards including Camera Link, Firewire and Gigabit Ethernet interface cameras that are becoming increasingly popular. With the proper combination of these technologies and a flexible architecture, a PMC module virtually eliminates the risk of obsolescence by ensuring backward and forward compatibility to keep systems performing at peak as technology advances and assures users maximum long-term return on their investment.
With ATCA blade architecture, integrators can easily integrate multiple cameras to an AVP for AOI utilization. For example, if an AVP blade processes image data from two camera link cameras, then 10 camera link cameras can be supported with a 5U ATCA system (Figure 1). Compared with the traditional sever solution, an ATCA-based AVP does not need five servers to achieve efficient performance. This type of compact form not only allows an ATCA-based AVP to save a considerable amount of space, but also helps simplify and reduce the large mesh of wires, connections and infrastructure that many industrial users deal with, thus freeing valuable physical space for other uses. The hot-swapping and full redundancy features of the ATCA platform also ensure that any component removal does not interrupt operation, as well as dramatically reduce maintenance, repair and replacement time.
Equipped with a base/fabric interface, an ATCA-based AVP enables system integrators to build communications between vision blades, including a Gigabit Ethernet base interface that provides a faster channel when exchanging image data among vision blades.
With a system management module mandatory on every ATCA system, SIs can easily build a highly available and manageable platform by installing a Chassis Management Module (CMM) to monitor and control the vision blades and other Field Replaceable Units (FRUs). The CMM watches over the health of the shelf, reports anomalies, takes corrective actions on errors, and interconnects across the system. The ATCA architecture’s reliability and flexibility are sure to improve machine vision operations thus allowing users an opportunity for substantial cost savings through the reduction of direct labor and improved product quality.
An example of a blade for use as an AOI server uses an FPGA camera link frame grabber PMC module instead of an AMC card and is based on the ATCA architecture for advanced vision applications. Such a platform provides computing for advanced image processing across many applications and industries. Supporting up to 16 Gbytes of DDR2 memory, this AVP blade also provides high-capacity memory throughput for processing and storing high-density image data. Such blade servers represent a shift away from traditional proprietary machine vision system integrators by alleviating the need for a DSP and utilizing the power and performance of the ATCA architecture. For AOI applications, an AVP blade must utilize a high-performance CPU to process data at an optimal level while performing complex morphological operations. With a wide memory bandwidth and capacity, AVP blade technology further aids users by efficiently storing large amounts of image data that can be accurately tracked and debugged if necessary.
Another essential item in building a cost-effective and space-saving platform for AVP is a comprehensive I/O density that provides greater flexibility and manageability. This will enable it to support up to four Camera Link connectors for dual-channel output, VGA, GPIO, LAN, and two USB ports on the front panel. The AVP blade backs a Camera Link or IEEE 1394 interface built on a PMC module and installed on two onboard 64-bit/133 MHz PCI-X slots. With an integrated CompactFlash card slot, developers can conveniently build an OS image to boot the system. A rear transition module provides an extra 2.5” SATA HDD storage that integrators need to store images for further analysis.
The onboard PMC module of the AVP blade for a high-performance Camera Link or IEEE 1394 interface supports high-speed image data transfer. Write/read wrappers around the FPGA manufacturer’s IP core were developed so that it can achieve the benefits of a high-capacity system, allow users to define the bus width of each write/read port, and enable calculation efficiency. Through the onboard high-density, customizable, FPGA-based processing core, each channel supports image data transfer rates of up to 640 Mbytes/s with an acquisition pixel clock rate of up to 85 MHz.
A standard platform requires only a customized PMC card for full functionality. Specialized AMC cards are not necessary. The platform shortens application development and system upgrades while allowing SIs to implement their proprietary FPGA know-how for pre- or post-image processing.
As manufacturing environments evolve, they produce items that are delicate or susceptible to contamination, such as integrated circuits and temperature-sensitive devices, as well as pharmaceutical products that may require measurement methods that are noncontact and nonintrusive. Vision inspection is also advantageous in processing applications where safety is a factor, such as parts made from hazardous materials. Properly fitted AOIs not only add value through improved efficiency in many highly technical processes, but also allow for on demand statistical collection and the ability for real-time feedback in manufacturing processes. The power, flexibility and connectivity of AVP systems have become almost necessary to achieve 100% inspection with high throughput at relatively low cost.