Research On Defect Detection Algorithm Of Solid Wood Plate Based On Deep Learning

Wood resources are limited in China.In order to improve wood utilization,machine vision is used to achieve fast and stable detection of wood defects.Not only can it overcome the disadvantages of low manual detection efficiency,high labor intensity,and low accuracy,but also have important significance for improving the level of intelligence of wood processing enterprises.This paper expounds the research status of machine vision inspection technology and equipment at home and abroad,and introduces the related theories and algorithms involved in the detection of wood defects based on machine vision.Machine vision is a simulation of human vision.Combining machine vision and artificial intelligence,it can greatly improve the speed,stability and accuracy of detection.It can greatly improve the detection speed,stability and accuracy,which has important theoretical significance and high practical value for accelerating the transformation and upgrading of the wood processing industry and improving the level of automation and intelligence of enterprises.Therefore,this topic puts forward the research of double-sided defect detection of solid wood plate based on machine vision.The research contents are as follows:(1)Firstly,the image acquisition device was built.Taking pine and fir wood as the research object,the plate specimens images contained major natural defects such as joints,dead joints,cracks,and decay were collected and saved to the sample database by the device.To improve the detection accuracy,the deep learning algorithm was used to detect the defects of solid wood plates.(2)A few classical object detection algorithms were analyzed,such as R-CNN(Region-Based Convolutional Neural Networks),Fast R-CNN(Fast Region-Based Convolutional Neural Networks),Faster R-CNN(Faster Region-Based Convolutional Neural Networks).The advantages and disadvantages of each algorithm were compared.Finally,the Faster R-CNN algorithm and SSD(Single Shot-multibox Detector)algorithm were selected as the research model for this subject.The algorithm and general processing flow of Faster R-CNN were introduced.Taking the surface defect image of solid wood plate as the object,the SSD algorithm was used to detect the defects and analyze the results,and it was found that the SSD algorithm was deficient in the recognition effect of small features and the detection ability of small target features was insufficient.Therefore,an improved SSD algorithm was proposed.Through the comparison of experimental results,it was found that Faster R-CNN and improved SSD algorithm had very superior performance for the detection of common defects on the surface of solid wood plates.(3)A defect detection system was developed for solid wood panels based on deep learning.The original image information table of solid wood board,the image information table of solid wood board defect,and the information table of classification level of solid wood board defect are established through SQL Server software tool.The interface of the solid wood plate defect detection system was designed and the functions of each module were introduced,it had good human-computer interaction.