Online Inspection Of Thin Film Key Line Switch Defects Based On Machine Vision

The membrane button circuit Switch is an important part of the keyboard,and is mainly used to fix keyboard keys and transfer key information.The membrane button circuit Switch is mainly composed of a button silicone cap attached to a flexible printed cir cuit board.Among them,the button silicone cap is an important part of the membrane button circuit Switch,and is mainly used for fixing the button.In the production process,due to the bonding quality of the button silicone cap and the material of the product,the membrane button circuit Switch often has quality problems.The membrane button circuit Switch currently relies on manual detection,which is costly and inefficient.In this thesis,a machine vision-based film butto n circuit Switch defect detection method is proposed to detect the sticking quality of the butto n silicone cap on the membrane button circuit Switch.This paper gives the design structure of the hardware system and the key algorithm of the software system,and gives the test results.The test results show that the scheme can quickly and accurately detect the sticking quality of the button silicone cap on the membrane button switch.The whole detection scheme is divided into two aspects: hardware design and software algorithm.In terms of hardware design,a line camera,a tunnel light source,and a telecentric lens are used to form an imaging system.A digital fiber sensor,motion control card,and conveyor belt are used to form a motion control system.In terms of software algorithms,on the one hand,MFC is used to establish human-computer interaction interface.On the other hand,different detection algorithms are written for different defect types.Among them,the gray-based template matching method is adopted for the off-button silicone cap;the broke n button is used.The silicone cap is judged by the method of line detection.For the sticky and reversed button silicone cap,the gray projection of the radial expansion of the button silicone cap template image is proposed as the feature vector,and the multi-layer perceptron neural network is used for training.And identification;for the multi-adhesive button silicone cap,the morphological method is used for screening.Finally,through a large number of actual tests at the industrial site,the average missed detection rate of the system is 5.31%,the false detection rate is 6.61%,and the single-chip detection time is less than 1200 ms,which basically meets the production design requirements.