Research On Glue Width Detection Of Reverse Osmosis Membrane Filter Based On Machine Vision

The safety of drinking water plays a vital role in human health,reverse osmosis water purifiers have become the primary choice for household water purification,and the core of a reverse osmosis water purifier is the reverse osmosis membrane element(cartridge).The process of producing reverse osmosis membrane elements mainly includes welding,gluing,coiling,cutting,etc.The gluing process is primarily to seal the three sides of the central tube,and the gluing process will have a wide or narrow gluing line that will affect the performance of the reverse osmosis membrane elements.If the gluing line is too little after the reverse osmosis membrane elements are cut,it will cause the seal to fail to produce water leakage,and the gluing line is too broad,it will cause If the glue line is too comprehensive,the area of the reverse osmosis membrane filtering water will be reduced,resulting in lower water production.Due to the particularity of the colloid itself,manual contact measurement will cause damage to the colloid,the measurement efficiency is low,and the measurement error is significant.In order to solve this problem,this paper proposes a non-contact detection method based on machine vision to realize the non-contact automatic detection of the width of the glue line.The current situation of the detection of the width of the glue line in the process flow of the enterprise is analyzed,and the demand for the size of the width of the glue line is put forward.The width of the glue line needs to be controlled in the range of 5 ± 0.5 mm during the operation of the reverse osmosis membrane element automatic film-making machine,and the overall detection time needs to be controlled within 2 s.The general scheme of the visual detection system for the width of the glue line is designed.Based on the existing reverse osmosis membrane element automatic film forming machine,an optical detection platform for the width of the glue line is built.The overall structure of the glue line width detection system is designed.The hardware required for the detection system is reasonably selected;the main modules of the detection system are also developed.The collected glue line image is preprocessed.Firstly,the original image is grayed,and the region of interest of the glue line is extracted.Secondly,the main source of noise is analyzed,and the extracted region of interest is filtered and denoised.Because the target glue line and the background diaphragm are low contrast,the image enhancement algorithm suitable for this low-contrast scene is selected through comparative analysis,and then the gray-scale morphological operation can be used to remove part of the reflection on the glue line.Aiming at the fact that the overall contour of the edge of the glue line needs to be continuous and occupies a pixel width,an improved Canny edge detection algorithm is proposed to locate the edge of the glue line roughly.On this basis,the edge of the glue line is precisely positioned.The coordinates obtained by the moment method and the polynomial interpolation method in the sub-pixel edge detection algorithm are compared and analyzed.The polynomial interpolation method with higher positioning accuracy is selected to extract the edge of the glue line.A set of algorithms has been developed for overall glue line width detection and for determining the quality of the glue line.The glue line width visual detection software has been designed and developed according to the functional requirements,and the reliability of the algorithms and the detection software has been experimentally verified.The checkerboard calibration board is used to calibrate the industrial camera to obtain the calibration coefficient.The whole glue line is divided into 273 small sections,and the detection rectangle is numbered in turn.The detection results of the glue line are too broad,and too narrow,and the regular part of the glue line is analyzed.The results of manual measurement and algorithm detection are compared.The results show that the detection error and detection efficiency of the system can meet the requirements.