Web Drives Vision Systems Specialization

Traditional web applications built around inspecting wide surfaces on products as diverse as paper and logs, to rice chutes, are among the most challenging machine vision applications in the field today.  These systems typically run at very high speeds, and unlike bottling lines that run at high speeds, the product is continuous, presenting new challenges in frame-to-frame registration and defect identification.

To meet the growing number of web applications in today’s challenging economic landscape; machine vision suppliers are evolving hardware and software to cost-effectively meet increasingly strenuous application requirements.  From lighting to smart camera platforms to software, all points of the machine vision compass lead to new solutions for web applications.

Manufactures Still Pay for Productivity
If you make millions of diapers per year, finding a way to save a millimeter of material on each diaper can save you millions of dollars.  That’s what Jim Anderson, machine vision product manager at SICK (Minneapolis, Minnesota) is seeing despite challenging macroeconomic times.  “People making consumer staples, the kind of products required for daily life, don’t slow down much; people have to buy their product,” explains SICK’s Anderson.  “If they can squeeze out an extra half of a percent of revenue, it becomes a viable solution with almost immediate return. That’s why we designed the AT20E for precise web position control.”

By moving to a cost effective CMOS sensor with specialized image processing ASIC for generating analog values on web position, and white light line generator rather than laser line generator, SICK’s new AT20E can locate an edge with less than 100 micron accuracy at a very cost-effective price compared to traditional retro-reflector or laser-based web tracking vision systems.  “Diapers aren’t the only staple, of course.  We’re seeing the same thing in comfort foods, like macaroni, for instance,” concludes Anderson.  “People are still buying these products.  Perhaps buying less, but that gives all the more reason for the manufacturers to cut waste and improve efficiency to lift the bottom line.”

An Enlightening Experience
Part of the flexibility of SICK’s laser-based Ranger and AT20E white-light web systems is the company’s use of CMOS area arrays that can be windows to a small area so that the system can work with less intense light sources, such as diffuse white light, rather than intense laser light.  But moving to white light illumination holds other benefits as well, such as the ability to image in color.

LMI Technologies (British Columbia, Canada) is using CMOS area arrays in lumber, tire, and road web applications with new lighting methods to enable new color imaging functionality while cutting down on overall system cost and footprint.  “LMI has been doing dimensional scanning and optimization of lumber for many years,” explains Barry Dashner, cofounder of LMI Technologies.  “The next stage is to go color or black-and-white for defect localization.  We chose to go with color through our FireSync solution about 2 years ago.”

By moving from linear array to CMOS area array, LMI can move the camera closer to the board or log under inspection, making it easier for OEMs to design the vision system into the manufacturing equipment.  The shorter the standoff between camera and surface lessens the light intensity requirements because an area array may work at hundreds of frames per second, while a linear array must operate at thousands of frames per second to acquire the same image; the shorter the exposure, the more light you need to get a good image pixel.  This has allowed LMI’s customers to move from high-intensity fluorescent lighting with high maintenance-associated costs, to strobed white LED solutions that are much more rugged and long lasting.  “Each time a light is changed, the company has to go through a pretty rigorous alignment and calibration routine to make sure the system is working properly.”

According to LMI’s Dashner, web applications continue to demand new, more powerful solutions while benefiting from simpler programming and operation. “If you’re looking at printed material, there’s a pattern that doesn’t change unless there’s a defect.  The goal is to have software that looks for these exceptions and only reports the exception, allowing the camera to handle more data at higher speeds.  We’re working hard to make our GUI’s and customer information more accessible.  With more technology and more complex applications to be effective, the systems need to be easily accessible and understood without needing a vision engineer nearby.”

Food Processing: Faster than Fast
Another way to beat the light budget issue while sticking with linear arrays is to go to the largest possible sensor pixel size.  Imaging Solutions Group’s (Fairport, New York) co-founder, Kerry Van Iseghem, tells how his company evaluated most linear arrays on the market before selecting a Perkin-Elmer linear array with 14x14 micron pixels.  “Most area arrays are under 3 microns, although Kodak does have some CCDs that are 7.4 microns – and that’s a big pixel for an area camera,” explains Van Iseghem. “But our line scan sensor has an active area that’s 4 times larger, so you’re going to have 4 times the sensitivity.  You can also put amplifiers, filters, etc., right next to every pixel, making these cameras an elegant solution for moving targets like rice sorting, coal conveyor inspection, and other applications.”

ISG offers a family of 1394 based machine vision cameras in addition to custom cameras for specific OEM applications.  “What separates us is the level of integration… We virtually eliminate host PCs and unique peripherals through the use of FPGA and other processors, delivering the same capability of a $10,000 vision system at a fraction of the price.”

Sorting food flying through a chute demands the highest machine vision performance power.  “With rice sorters, the grains are flying by and you have to scan 4 tons of rice an hour and then direct a bank of air jets to shoot out the discolored rice, rocks, etc., in mid-air within 2 inches of passing the camera,” explains ISG’s Van Iseghem.  “The tricky part is getting the timing right, but once you’ve done that, these sorting systems are extremely accurate and fast… on the order of a single or double-digit microseconds.” ISG has used similar line scan technology to help companies in Canada with automated road inspection conducted from moving vehicles as well as monitoring 17-mile long coal conveyors deep below the ground.

Coatings and Panels
A slowing economy may hurt flat panel sales for a short while, but the good news is the growing number of coatings and applications for flat panel applications that require precision machine vision to identify and locate defects.

Taiwan’s Winstar Technology is an integrator that uses EURESYS Grablink frame grabbers and eVision image processing library to meet quality inspection requirements in plastic film, paper, metal foil, optical film and protective hygiene plastic barriers, as well as flat panel inspection.

“The new applications [in Asia] are optical film and flat panel display inspection…[which represent] about 85% of our focus in Japan, Taiwan, Korea, and some in China,” explains Winstar’s General Manager, Alex Liao.  “The process includes surface defect inspection in color filter, mother glass, and PMMA [Polymethyl methacrylate] by sheet, and some polarizing film, light diffusing film bright enhancement film, etc., in roll to roll production.”

This keeps cost down while continually moving to larger flat panels requiring ever-larger coatings.  The industry continually seeks CCD cameras with higher resolution and higher speed.  “Normally, camera rates were at 40 MHz, but they have progressed to 60, 160, and 320 MHz with defects size as small as 20-45 micron.  These defects are barely detectible by human eyes, but in flat panels, the defects such as particles, micro-gel, dirt, pin-hole, and scratches are magnified 10 times by optical effects, so the defect becomes unbearable.”

Line scan cameras have dominated web applications for many years because of the nature of the moving target, and line scan’s ability to create continuous images by stacking one line of pixels on top of the next.  However, this approach places extreme requirements on the lighting system to keep pace with the growing speed of web inspection systems.  In response, vision manufacturers are releasing new designs that use windowed CMOS sensor that adapt to the speed and light budget of the system along with new processing elements and algorithms to lighten the computational load on host PCs, or eliminate the PC all together for easier OEM integration.