ATTN HPCN-TTN
[DEUTSCHE Projektbeschreibung]

The HPCN-WOOD Project


HPCN Awareness Campaign about Application of Digital Image Processing for Grading in Wood Industry

The reduction of production costs is an essential necessity to operate on a market, where more and more products come to the EU from countries with low labor costs and low prices for raw material. There are two principle ways to reduce the costs: First, speeding up the whole production line. And second, to replace expensive labor by machines.

These were basic assumptions of the HPCN-WOOD project and were pretty much verified by the evaluation of a questionnaire, which was part of the project.

The results show, that increasing the efficiency and getting automatic documentation and quality assurance are the most expected advantages. There is a rather high amount of doubt, that computer based optoelectronical grading systems can really replace human graders. It has to be demonstrated by really established systems, where the demands are comparable for a possible future user.

Speed increasement reduces production costs

Speeding up a production line can reduce the production costs of each piece, if most of the production line is able to run at higher speeds without the need of expensive modifications. Today very often the grading task is the bottleneck in the production line. Grading in wood industry means, to scan the surface of the piece of wood at rather high resolutions to detect even tiny defect like worm holes or cracks. The resolution, necessary for this tasks are sometimes better than 0,2 mm/Pixel. Production speeds are increased to more than 4m/s. Grading wooden surfaces correctly with better results then human experts is a very complex task. No region on the natural grown wood surface looks exactly like any other region, and now defect (like a barkringed knot or a crack) looks exactly like any other defect. The algorithms for classification also have to be sophisticated to be tolerant to natural variations of the input material, but highly sensitive to even tiny serious defects. Therefore, very high data rates of up to 20 MHz/ camera input data have to be handled under real -time conditions. Simple PC solutions are not capable to perform the complex algorithms, necessary to describe wooden surfaces and discriminate between different regions at this high data rates. A network of DSPs (digital signal processor) is a good solution to meet the requested high classification accuracy and the real time capability.

A very important advantage of automatic grading systems is a high and stable rate of correct classification results. Well trained graders, even if they work with all their attention because they are supervised during an assessment, can not really achieve a high repeating accuracy. The diagram on the left side shows a result of manually grading strips of a parquet floor into 3 different classes. Only slightly more than 40 % of the strips were classified 4 times into the same class. A well designed automatic system can guaranty a stable high rate of correct classifications of better than 85 %.

 Accurany of manual grading

The costs for installing or upgrading an existing production line must be divided into two parts. Improving or constructing the mechanical part for handling the goods, and improving or supplying the optoelectronical system. The costs for higher speeds are rising only slightly for the optoelectronical system. An optimum solution can be found, where the costs for improving the mechanization are not increasing too much to make the whole system too expensive.

For the assessed application of increasing the speed of an existing black and white parquet grading system by a faster color system, an amortization time of less than 1,5 years can be reached. In addition, several limitations of the old system, like the detection of sapwood and color defects can be overcome by a new HPC based DSP color system. The other expected advantages, like automatic documentation, quality assurance and flexibility can be achieved easily by such a system.

This activity is associated with the ATTN technology transfer node.

For more information on HPCN-WOOD have a look at the page from Joanneum Research.