Identification of Timber Defects Using Convolution Neural Network

Timber is important in the construction industry. Random checking in the manufacturing line, randomized product selection, and condition monitoring are some of the most common quality control methods for wood panels. Manufacturing can continue if the checked products' expectations are met; however, if flaws or anomalies are discovered, the supply chain is shut down, a remedy sought, and the manufacturer loses productivity. Skilled human operators continue to perform the bulk of visual quality inspections in the timber products industrial processes. Visual inspection is a time-consuming and labor-intensive process that is prone to human error. New automation processes based on high-resolution images and inspection systems are currently being developed, but these are yet quick or not accurate enough for real-time industrial processes. Using an automated visual inspection device, the whole project proposes a method for recognizing and classifying defects on the surface of wood. Here, a convolution neural network has been used to detect the defects present on wooden surface layers.