Optical sensor helps to speed up car production

05 May 2005

French firm ActiCM's high-speed, vision-based co-ordinate measuring machine is grabbing the attention of big names in the car industry. James Tyrrell finds out why.

FromOpto & Laser EuropeMay 2005

ActiCM's robot-mounted imaging system promises to slash car body-inspection times by a factor of five or more. French car-maker Renault is so impressed with the company's vision-based co-ordinate measuring machine (CMM) that it plans to purchase a system in 2005.

"Renault's problem is like that of every car manufacturer using traditional mechanical CMMs," Antonio Mendes Nazare of ActiCM told OLE. "To inspect and validate an assembled car body involves measuring about 2000 elements. With a traditional mechanical CMM, this takes about 10 h."

Considering that Renault was applying a series of mechanical, probe-based measurements to every one of the thousands of vehicles rolling off its production line, it is clear that the car-maker has a great deal to gain from a faster way of validating car body shape. "[About two years ago] they launched a benchmarking process with all the big names in CMM to try and find some other solutions," explained Mendes Nazare. "The firms came to Renault and demonstrated what they had to sell, but none of them were satisfactory."

At that time, ActiCM, which was founded in 2000 by two former French Atomic Energy Commission (CEA) engineers, was still developing its vision-based alternative. From its Moirans location near Grenoble, France, the firm was busy investigating the idea of combining digital-image processing with photogrammetry - a technique for extracting 3D co-ordinates from 2D images.

ActiCM was so confident in its new technology that it took its optical sensor concept to Renault. Interested, the car-maker asked the firm to assemble a pilot system that could be put through its paces at the plant.

As part of its benchmarking process, Renault requested that machines deliver an accuracy of 300 μm over the whole car body on the production-line site. "This is important because [although] the big CMMs can achieve 50 μm, when you put them in a factory environment they are not better than 300 μm," said Mendes Nazare. "With a vision-based non-contact system we can achieve the same accuracy as traditional CMMs on the whole volume of the car body, but much, much faster."

Mendes Nazare boasts that ActiCM's pilot system has cut the time taken to perform the 2000-element validation from 10 h down to 2 h, with multiple sensors offering further reductions. "With four sensors they [car-makers] can get down to 30 min for the whole car measurement," he explained, "20 times faster than conventional CMMs."

System details Placed on a robot, up to eight of ActiCM's sensors can work together simultaneously on the same part. The first step is to capture an image and the second is to process this image to extract 3D co-ordinate information. To validate the part, these co-ordinates are referred to a computer-aided-design file and compared with original specifications.

Dubbed an intelligent vision device (IVD), each sensor unit has two high-definition (multi-megapixel) CCD cameras, a visible LED illumination source and a white-light projector. The system can also operate under ambient lighting conditions if they provide sufficient contrast. "It is called an intelligent vision device because it knows what light source to use," said Mendes Nazare. "The system switches between different lights depending on what it has to measure."

The IVD's onboard cameras are able to image a part from two different and known positions. These two views can then be geometrically manipulated by the device software using photogrammetry, a method which dates back to the 1850s, to extract 3D data.

"Today, photogrammetry is a well-known technique, but actually it was not used very much until 15-20 years ago because of the amount of calculation it requires," commented Mendes Nazare. "Only with increases of computing power has it become possible to use it in industry."

Advantages Although the IVD is robot-mounted, the mechanical arm is simply there to guide the sensor around the car body and enable an initial camera calibration. "The robot is really just a way of transporting the IVD. I could [in principle] guide the IVD by hand," said Mendes Nazare.

The sensor's CCD images provide all necessary distance and depth information. This is a big advantage as it means that the accuracy of the device, unlike many conventional CMMs, is not related to the positional accuracy of the robot. What's more, an imaging-based approach is much faster.

"A [camera] picture covers 1 x 1 m, and every element within this picture will be measured," explained Mendes Nazare. "This means that, if you have 50 elements within the picture, in around a second you will have 50 acquisitions, which is why our system is so fast."

Imaging technology offers further advantages over conventional CMMs when it comes to re-examining data. "When you have a CMM, you have to make a sequence tree," said Mendes Nazare. "You have to tell the system what elements you want to measure." In effect, once the car-body measurement has been performed, co-ordinate data will exist only for elements that have been specified within the tree, unless you happen to be using ActiCM's device.

The firm's "all-seeing" CMM allows designers and engineers to go back to the sensor's original images and reprocess them to measure elements missing from the original sequence tree. "This is something very new [for the industry], because, previously, if you didn't think of measuring something at the time, it was lost," explained Mendes Nazare. "Now, even if you think an element is unimportant [today] you can still find its position [tomorrow] on the image library."

Attracting interest With such a powerful system in its hands, it is no surprise that ActiCM has attracted interest from a number of major automotive firms. Although working closely with Renault, ActiCM also has relationships with Hyundi, Toyota and Nissan. Mendes Nazare stresses that every car-manufacturing plant is a potential deployment site for the company's vision-based technology.

"We are manufacturing two systems," said Mendes Nazare, "a fully portable unit [Acturis] and a second high-speed non-contact CMM [Advent] for production-line control and inspection."

The firm has taken great strides to commercialize the technology from its origins at the CEA. "It took about three or four years to industrialize, because at the CEA it was mainly theory," revealed Mendes Nazare. "We came up with the first product [Acurist] at the end of 2003 and, at the same time, we were ready to start the Advent, which is when Renault purchased the pilot."

The pilot system took ActiCM one year to make and was delivered to Renault last year, passing its validation in December. Mendes Nazare estimates that, depending on the nature of the application and the measurement speed required, the purchase price of its Advent system will be somewhere within the €200,000-500,000 range.

"The first application was for car-body measurement, but now we have requests from manufacturers for doors and bumpers," Mendes Nazare said. "[Outside of the automotive sector] we have had requests from the aerospace industry."