The largest book distributor in Germany packages more than 6500 parcels per hour. To cope with such throughput, the company uses an automated mailing system that automatically monitors every package via attached barcodes containing customer information. As each package travels though the automated mailing system, each barcode is examined, and a relevant mailing label is generated and affixed.
System developer Fränz & Jaeger GmbH (Aachen, Germany; www.fraenz-jaeger.de) was contracted to develop inspection equipment that could process approximately 85% of these packages automatically. In the book distributor's facility, an Ethernet-based system using an array of barcode readers from Sick (Düsseldorf, Germany; www.sick.com) was already in use.
null
According to Thomas Fränz, managing director of Fränz & Jaeger, these scanners can detect approximately 75% of the barcode information contained on the packages. To improve this detection rate, a more powerful barcode-reader system was needed to compliment the scanner-based units. Fränz & Jaeger developed an OCR reader using off-the-shelf cameras, frame grabbers, lighting, and computer systems that enable the final system to detect up to 85% of the data on the packages.
"As parcels move under the system at greater than 30 m/min," says Fränz, "high illumination was needed to capture images from the packages." To do this, a linear fiberoptic illuminator from Crescent Lighting (Thatcham, Berkshire, UK; www.crescent.co.uk) produces 120,000 lux at the face of the parcels.
To take advantage of this illumination, a Pirahna linescan camera from Dalsa (Waterloo, ON, Canada; www.dalsa.com) is mounted on a vertical motorized stage positioned approximately 50 cm above the conveyor belt containing the parcels. Using a 60-mm lens from Linos Photonics (Göttingen, Germany; www.linos.com), the camera can image a field of view of approximately 60 cm across the width of the conveyor.
"Because packages entering the system range in height from 3 to 50 cm," says Fränz, "the camera system is mounted on a motorized stage along with an RS-232-based laser scanner." After the laser scanner measures the distance from the camera to the surface of the package, the imaging data are transmitted to a host PC and an RS-232-based stepper-motor controller that adjusts the height of the stage to a positional accuracy of 0.15 mm.
The linescanned images are captured into an Ethernet-based PC using a PC-DIG frame grabber from Coreco Imaging (St. Laurent, QC, Canada; www.imaging.com). Running under Linux OS from SuSE (Oakland, CA; www.suse.com), the PC analyzes the images using the picCOLOR modular scientific and industrial image-processing and analysis software from Fibus (Düsseldorf, Germany; www.rhein-ruhr.de/~fibus/).
"OCR/OCV was implemented in picCOLOR by using a regression pattern-matching approach," says Fränz. "The most interesting function of this module is a character-orientation analysis. During operation, picCOLOR recognizes the orientation of lines of text and properly rotates the text into correct alignment before the character analysis." OmniPage OCR software from ScanSoft (Peabody, MA; www.scansoft.com) can also directly be addressed from picCOLOR and enables the system to accomplish more complex reading tasks.
Once the system has recognized the label on the package, the host CPU transmits the data over an Ethernet interface to a central CPU. Here, customer information is extracted from barcode information and a label is generated using a Zebra high-speed printer from Etimark (Bad Nauheim, Germany; www.etimark.de). After the label is automatically attached, the parcel is routed to the shipping department for final mailing.
"Incorporating the system with the previously installed barcode readers allows 85% of packages to be shipped automatically," says Fränz. "Packages that cannot be read by both systems are routed to manual checking stations, where the data are hand-keyed into the host computer system."
