On the suitability of colour and texture analysis for detecting the presence of bark on a log

• Three algorithms for distinguishing bark and wood on spruce log images are compared.
• The algorithms are based on colour, texture and combined information.
• Influences on the performance of the algorithms are identified.
• The colour information per pixel is not sufficient on its own.
• The texture algorithm and the combined algorithm show promising results.

This paper deals with pattern recognition based on colour and texture. The aim is to detect areas of bark and wood on images of spruce logs taken in the industrial environment of a sawmill. Bark is a material with high natural variability in texture and colour even in a single tree species such as spruce and is a challenging subject for pattern recognition. The study compares three algorithms based on colour, texture and an integrated combination of colour and texture, respectively. Relevant sources of variation between and within logs are identified: colour variations on bark and wood, dirt, and traces of the bark beetle. Modifications of the algorithms with the potential to further improve the bark detection performance are discussed.