Machine Vision News
Vol. 11, 2006
Vision Club of Finland
Previous
Index
Next

From prototype to product – case x-ray inspection

Atostek Ltd. expertises your R&D project. From our perspective, implementing a system proto in the beginning of the R&D project leads to good results in finding the right technical features for the final product. ’From prototype to product’approach is demonstrated in the following case of x-ray inspection.

This article presents a machine vision system that is based on an x-ray device. The system is used by a manufacturing company, who wanted to inspect manually assembled product packages. The development of the system started from a prototype implementation and after a period of pilot test, it was delivered to the manufacturer by Atostek’s customer. In this project Atostek did technical studies, system design and software design tasks, and implemented the image analysis and user interface parts of the final software. Atostek’s customer took care of the assembly and delivery of the system.

Project objectives

The manufacturing company wanted to monitor the quality of the manually assembled packages. Additionally, they wanted to catch human errors like missing components. Traditional methods that are based on weighting the product are not reliable enough since the components are getting smaller and lighter. An x-ray monitoring system is expected to reduce the number of misassembled packages that are shipped from the factory.
The operation of the system was required to be traceable. An image of every box and a box ID had to be saved in order to handle possible reclamations. Components may get lost at every stage of business, including the distribution chain.

Parts

The inspection system (Figure 1) consists of conveyor belts, x-ray device, barcode reader, failed box pusher,analysis computer, and database server


Figure 1. System architecture

Operation

The operation of the package inspection system is shown in Figure 2. The packages are packed in boxes of five, which arrive to the x-ray device from the packaging cell. The barcode reader reads the box ID, and the x-ray device takes a picture of the box. The picture is classified by the analysis computer in real-time and, according to the classification result, the box is passed thru the system, or it is pushed to the fail conveyor. Finally, the picture is saved to the database along with the classification result.


Figure 2. Device on the factory floor

An operator sees the classification results of the packages on the user interface (Figure 3) as the packages are classified. The missing components are highlighted for the operator, as shown in the Figure.


Figure 3. Main window of the system. Missing components in the second package

Teaching of the inspection system

It was clear from the start that traditional, figure based shape detection would not work in this application, since the components can be on top of each other, they can create different kinds of shapes, they can be in several different positions, and some of them are fairly small.

There are several hundreds of variants of the packages, which mean that the teaching of a new variant should be as easy and as fast as possible (Figure 4). An innovative teaching method made it possible to teach the new variants while the packaging line was still operational. It is also possible to dynamically fine-tune the teaching parameters during normal operation. Teaching of a new package variant takes about 15 to 30 minutes.


Figure 4. Teacing window of the system

Design challenges

All the features discussed earlier were decided to be implemented in the prototype. The main challenges for the prototype implementation were:

  • A tight schedule
  • Integration of the x-ray device to the system
  • Recognition of the components from the package
  • The system configuration was required to be easy

One month from concept to the pilot testing

It was important to get rapidly to the pilot test phase, since a large enough flow of packages and different package variants could not be tested under laboratory conditions. Due to this, the real cycle time of the system and the reliability of the classification could only be verified in a real operation environment.

The pilot test phase yielded a large image database. This database was used to improve the classification algorithm. In addition, manufacturer’s final requirements for the system were resolved.

Improvement of the system was not the only reason for the pilot testing phase, but it had also some business-oriented motives. Feedback from the pilot testing phase made it possible for Atostek’s customer to choose the R&D ideas that best fit to manufacturer’s needs, and enabled the manufacturer to develop his processes in order to integrate the new device into the factory line.

Conclusions

During the pilot test phase the system was developed further and now it has been delivered to several locations around the world. The system can be used in different lines of industry, for example, food industry, medical and electronics wholesale.

This article discussed an x-ray inspection system that was developed for Atostek’s customer. Atostek added value to the project with technical consulting, software design, implementation and maintenance. Prototype implementation played an important role in the development of the system. It is essential to understand that instrument design is not enough in modern devices but IT design is needed. This is the area in which Atostek is more than willing to help you.

Contact Information:


Atostek Oy
Mr. Jani Sarviluoma
+358 50 596 3848
jani.sarviluoma@atostek.com
www.atostek.com
Hermiankatu 8 D
FIN-33720 Tampere
Finland

Previous
Index
Next