Stand-alone sensor takes over systems tasks

Until now, optical character recognition has been the preserve of PC-based vision systems.

Until now, optical character recognition (OCR), a method of detecting and reading characters, has been the preserve of PC-based vision systems.

The deployment of the In-Sight 3000 vision sensor as a stand-alone system at ZF Getriebe in Saarbrucken has made this application simpler and more economical.

ZF Friedrichshafen is the largest independent supplier of automotive driveline and chassis technology worldwide, with about 55,000 employees at 117 facilities in 22 countries.

ZF Getriebe, part of the ZF Group, is based in Saarbrucken, Germany and manufactures automatic transmissions for some 20 car manufacturers throughout the world.

The entire production process at ZF Getriebe Saarbrucken is controlled by a rigorous quality management system.

Machine vision processing supports production in several areas.

In the past a "type plate" showing the serial and parts list numbers was manually attached to each transmission by riveting.

This operation subjects employees to a high level of noise.

The test-bed planning department accepted the task of automating the "type plate" process.

This department is responsible for designing and equipping the final acceptance testing operation at ZF car driveline technology in Saarbrucken.

The new concept envisaged replacing the "type plate" with cast lettering and in a rectangular area on the transmission housing.

The variable transmission data (serial and parts list numbers) were to be directly integrated in the cast housing.

This would provide a permanent and reliable means of identifying each transmission.

Once marked in this way, cameras and image processing techniques could verify both the existence and the unambiguous legibility of the lettering.

The planning engineer responsible at ZF Saarbrucken (ZFS), Martin Schuler, remarks, "In this way we planned to simplify the process and enhance its efficiency while improving quality at the same time".

Once the requirements were defined and after careful initial planning by ZFS, a feasibility study was conducted, and the contract to implement the project was awarded to Joachim Richter, a company based in Konken.

This small enterprise, which employs 25 people, supplies flexible identification systems comprising automatic engraving, lettering and marking devices.

Besides offering standard products, the company is focusing more and more on complete custom solutions.

An inspection and engraving system with the following attributes was scheduled to be installed on the production line and integrated into the process data exchange network: exact detection of the inscription surface's position on the transmission by means of machine vision processing; and verification of the engraved lettering's flawless quality with high operational reliability by means of OCR.

The project was implemented and the hardware installed in August 2002, and considerable flexibility was demanded of the image processing system.

Only one transmission housing model was available for the initial laboratory phase.

In order to achieve the desired versatility for various transmission types, the system had to be able to accept continuous minor modification, which made the requirements even tougher.

The project also had to remain within an economical budget framework.

Richter's Oliver Bill, the project manager, recalls, "In the form of a stand-alone system, the Vision Sensor represented a space-saving solution that was easy to use".

It was good fortune that Cognex equipped its In-Sight 3000 stand-alone vision sensors with integrated OCV/OCR software as standard just before the project planning phase.

This played a key role in the award of the contract in the face of competition from two rival bids.

At low cost and with the benefits of simple operation and installation as well as a high performance capability, the In-Sight vision sensor can be managed, monitored and controlled at each stage of production from anywhere within the corporate network thanks to the integrated Ethernet interface.

Each vision sensor comprises an extensive pre-installed library of Cognex's market-leading vision tools.

Installation, use and parameterisation are extremely easy.

In-Sight 3000 offers maximum image processing performance in situ; it consists of an encapsulated high-performance processing unit with interfaces, a small progressive-scan digital camera and a control pad.

The application is very simple to set up and program by means of the control pad and a spreadsheet interface.

The vision sensor thus allowed the implementation of a solution that had previously been the preserve of PC-based systems.

The various transmission types arrive at the inspection and marking station in a random sequence.

The type plate surfaces on the transmission housings differ slightly because of natural variation in the casting process and the use of various outside suppliers.

The benefits of Cognex's position indication software come to the fore in such an environment, as regards both simple programming and reliability.

The program developed via the spreadsheet interface comprises a series of layers or levels.

First, the program seeks to detect certain edges and properties of the type plate surface.

If these are not unambiguously found, a search operation for other characteristics is automatically performed at the next program level.

When making adjustments to deal with additional requirements, Richter benefited greatly from the prompt response from the vision experts in Cognex's local German office.

The system makes sure that the engravings, including the serial number, are positioned exactly in the correct place on each casting.

Optical character recognition is one of the most demanding image processing applications, requiring a very high performance processor to run the complex evaluation algorithms.

This explains why OCR applications have previously been run exclusively by PC-based systems.

Compared with verifying the legibility of learned characters by OCV, an OCR inspection evaluates each character by referencing the alphabet and number system, which increases the volume of work by a factor of 40.

The characters produced by the automatic engraver must be unambiguously legible within a strict, adjustable quality band.

The vision system detects engravings that are too shallow or too deep, and thus difficult to read, with absolute dependability.

The character evaluation system also detects wear or damage to the engraving tool, and issues a fault message accordingly.

Five different transmission types are currently being marked and inspected in three shifts, and a further three transmission types are to be added soon.

In a test series of 1000 Jaguar transmissions that passed through the system, the vision processing system worked 100% correctly.

In this application the In-Sight 3000 vision sensor provides a reliable solution to the tasks of position indication and optical character recognition.

The successful outcome of this project has encouraged ZF Getriebe to carry out further project studies on additional applications for vision sensors within their manufacturing operations.

Back to top