Integrated defect detection and optimization for cross cutting of wooden boards

Cross cut operations in the secondary wood industry are an important industrial process. The cutting is based on the contents of different defects on the wood surface and automatic defect detection based on image analysis are now being used. There are two important aspects of this industrial process. Firstly, it is a need to formulate a mathematical model that is capable of taking all information into account, and secondly that the image analysis together with the optimization is performed in real time. We propose an approach where these aspects are included and where the image analysis and optimization are truly integrated. The image analysis in fact generates the mathematical model directly in which only feasible cuttings are included. This contrasts the existing system where an intermediate step of establishing feasible cuttings is required. An important and difficult extension to standard problem using fixed-length products is continuous-length products. We describe an approach which maintains a real time capability and still guarantees optimal solutions. We present numerical results obtained from real production data at a Swedish window manufacturer.