Sensors lower cost of simple inspections

In many machine-vision applications, it is not necessary to deploy complex products that perform high-speed imaging, pattern recognition, or morphological analysis.

Jun 1st, 2004
Th 151321

In many machine-vision applications, it is not necessary to deploy complex products that perform high-speed imaging, pattern recognition, or morphological analysis. Factory-automation systems, for example, may merely require systems to analyze the absence/presence of an object—its size, color, or position. In these applications, low-cost image sensors are replacing PC-based camera and frame-grabber combinations, especially in systems where high-speed, high-performance imaging is not a requirement.

"In analyzing the machine-vision market," says Kenji Yoshida, general manager of Aromat (New Providence, NJ, USA;, "we discovered that approximately 80% of machine-vision applications only require simple tasks to be accomplished and these can be implemented in relatively few imaging algorithms." Recognizing this trend, a number of companies, including Banner Engineering (Minneapolis, MN, USA;, Cognex (Natick, MA, USA;, DVT (Duluth, GA, USA;, ipd (Billerica, MA, USA;, Keyence (Woodcliff Lake, NJ, USA; www., and Mitsubishi Electric (Cypress, CA, USA;, have introduced low-cost image sensors that bundle cameras, machine-vision systems, software, and processors into small easily deployable packages.

Aromat will soon up the ante in low-cost machine-vision systems with the introduction of its LightPix family, a range of products that specifically target color, edge, and size detection. What makes the products noteworthy, perhaps, is not their functional specifications but their price. Each of the four sensors, whether targeted toward color, edge detection, or width measurement, incorporates a 352 × 288 CMOS sensor, white LED lights, on-board processor, serial RS-485 port, and digital I/O and is supplied with a pushbutton operation unit to initially train the device for a specific function. To perform machine vision using the sensor, the device must be situated at a distance of less than 4.5 in. from the part to be inspected.

Machine-vision sensors costing less than $1000 automate screen-printing alignment systems at M&R, a large printing-press manufacturer. In operation, fiducal data are located and confirmed across five screen presses and controlled using a touch-screen interface.
Click here to enlarge image


"To reduce the price of these systems to $1500 per sensor and $600 (in single quantities) for the viewfinder and programmable controller," says Yoshida, "we have leveraged the fact that Aromat is part of Matsushita/Panasonic Group, one of the largest suppliers of LEDs, embedded processors, and CMOS images." In OEM quantities, this price drops another 30%. Samples of the sensors are also being made available to large systems integrators on a 30-day free trial. Already, a number of vendors have incorporated the machine-vision sensor into their products.

In a process similar to CMYK printing, silk-screen printing requires registration lines to be marked on a number of silk screens so that when each printing occurs, each ink is properly registered. In the past, this process was performed manually and subject to human error. By incorporating two ANE101 LightPix image sensors located approximately 105 mm over each press printing station, one of the world's largest manufacturers of screen-printing equipment for the graphics and textile industries, M&R Printing Equipment (Glen Ellyn, IL, USA;, has developed several systems that automate the x-y registration process. To control this process on a five-color silk-screen production system, five press stations must be networked as the material passes from one stage to the next.

M&R interfaced each of the two sensors on each print machine head with two ANE101 sensors to a Mitsubishi Electric Q-series slave programmable logic controller (PLC). This was then networked with four more print machines to a master PLC over CC-Link, a field network system that processes both control data and information to provide efficient, integrated factory and process automation (for information see

To control the system, the PLC was interfaced to a GOT touch screen, also from Mitsubishi. "While the ANE101 system did not directly support fiducial recognition," says Yoshida, "the customer's demands for a number of automated systems made it worthwhile to develop an algorithm specifically for this purpose."

More in Image sensors