Research On Visible Defect Detection System Of MDF

Medium-density fiberboard is a kind of artificial board made of small-diameter logs,harvesting,processing residues and non-wood plant fibers as raw materials,urea-formaldehyde resin or other suitable adhesives are applied through slicing,cooking and other processes,and then hot pressed.Play an important role in the furniture and construction industries.After investigation,at present,the appearance inspection of the MDF products before leaving the factory mainly adopts the manual inspection method.This method has a complicated operation process,low inspection efficiency,and a missed detection rate of about 5%.Based on the above-mentioned drawbacks of manual testing,this paper designed a set of online inspection system for the appearance defects of medium density fiberboard according to the quality inspection requirements of medium density fiberboard manufacturers and combined with machine vision inspection methods.The paper gives a brief introduction and analysis of the types and causes of appearance defects of MDF;17 kinds of defects involved in the subject,including sand leakage,rough board surface,dust spots,coarse fibers,three-layer structure,sand trench,low strength,impurities,Water spots,oil spots,corners,chipping,burrs,dark spots,sand pits,sawing paste and saw marks for feature analysis,on the premise of meeting the requirements of enterprise quality inspection,design the visual acquisition system,mainly Including the reasonable choice of camera,lens,light source and lighting method,design of transmission system,etc.Collect images that meet the requirements of the inspection algorithm to build a sample defect library.In the defect detection algorithm module,the positioning of the detection target is achieved by fixed threshold segmentation and morphological methods;for spot defects,the commonly used threshold segmentation algorithms are compared,and the local dynamic threshold segmentation method is used to achieve the detection requirements;for sand leakage and board surface Roughly gives a gray-scale contour segment detection method;for dust spots and coarse fiber defects,an image enhancement method using median filtering for background estimation is given;for defects involving density detection such as three-layer structure,in the system introduce laser ranging sensor to find the relationship between distance fluctuation and density defects;use the Canny algorithm to extract the defects that need edge detection;use the Gaussian convolution to enhance the contrast for the defects of sand trenches and pits,use Gaussian Line detection method to detect.The system uses the VS platform to complete the system architecture,human-computer interaction interface design,and fiberboard motion control.This subject uses a multi-threaded parallel working mode to build a sample experiment model.The model system runs smoothly.Testing 18 samples can effectively distinguish 17 kinds of defects,and no missing or false detection occurs.