Peeled Veneer Detection System Based On Machine Vision

Plywood grade was decided by the surface quality of rotary-veneer. At present, the inspection of rotary-veneer is done manually, and the inspection result is used to cut out and grade of veneer, this reduces production efficiency and adds costs. Artificial intelligence technology is introduced into production; this can help to improve the automation in the plywood industry. This paper will base on image processing methods and artificial intelligence technology, and combining the features of rotary-veneer, to perform rotary-veneer surface defective images recognition and classification effectively. At last, a detection system of rotary-veneer defects recognition and classification based on national standard will get.The main work included in the thesis is shown as follows:1. Based on analyze of various rotary-veneer surface defect image, a series of effective improved algorithms of image enhancements and image segmentations are presented, and the detect objects are detected and labeled according above.2. Various defects features of shape, color and texture were analyzed, and introduce feature parameters based on knot defective image color accumulative histogram percentile distribution features. Thus, the defects feature set was provided can be classified accurately in defect recognition and classification.3. Recognition and classification is done manually, the big errors can be made on subjective think and background knowledge of operator. Therefore, The self-organization mapping (SOM) network as classifier with non-supervision was introduced.we choose image color accumulative histogram percentile distribution features as feature set, which can describe the characteristics of knot effectively. The G-SOM software was effective in recognition and classification of knot defects, and can get a perfect result.4. The texture of rotary-veneer can interference in defect detects, this paper presented a modified Fuzzy C-Mean algorithm (FCM). The approach of this algorithm is sample density of inter-class and distances of intra-class as comprehensive parameters, thereby to obtain the validity initial cluster centers. This algorithm can detect the texture and defects on rotary-veneer surface exactly.5. Based on machine vision rotary-veneer detect system was establish, On system, the experiments, such as CCD camera calibration, test of driving device, detect image processing, include recognition and classification, and parameter calculations, are performed. Therefore feasibility and validation is verified through experiments.