Edge-glued wooden panel defect detection using deep learning

Abstract The wood-based furniture manufacturing industries prioritize quality of production to meet higher market demands. Identifying various types edge-glued wooden panel defects are a challenge for human worker or camera. Several studies have shown that the detection with low, high, normal, overlong, short is identified but residue and bluntness highly challenging. Thus, present model identifies by computer vision and/or deep learning, whereas learning based decide pass having better performance. goal this paper provide an improved defect solution process automation. Therefore, system was designed takes input images from camera as raw image laser-aligned panel. automation then performs vision-based features extraction detection. aim solve problems using design implementation detection, can be stated (WDD-DL) artificial intelligence Automated Optical Inspection (AOI) consolidation. Possibly there exist several on edges while edge-banding in manufacturing. scope achieve accuracy feature algorithms final result classification WDD-DL AOI. uses Gabor, Harris corner, morphology, structured light curvature calculation pre-processing InceptionResnetV2 Convolutional Neural Network algorithm attain best results. applications work found control industry edge, joint panels. achieves results precision, recall F1 score 0.97, 0.90 0.92, respectively. experiments demonstrate achievement compared other methods overkill escape rate analysis. Ultimately, discussion section provides interesting experience sharing about necessary factors implementing real-time industrial operations.