In this paper a multisensor system for grading of timber in sawmill environment is presented. A fast X-ray sensor is used for the detection and the measurement of all kinds of knots in softwood. Because the X-ray system is independent from the appearance of the wood surface, an extremely low rate of false positive classification, caused by dust or decay for example, can be achieved. A four side black and white optical system distinguishes between the different subclasses of knots and classifies all other recognizable wood defects. For the optical system a texture feature based algorithm with pixelwise classification was chosen. The classification is done by a hierarchical classification process with adaptive decision rules, depending on the overall quality of the actual board. The system was installed in a sawmill in Italy in autumn 1996.