| The lower tool part is one of the important parts of the sewing machine.The lower tool part and the upper tool part together form the tangential module of the sewing machine,which is often used in the tangential work at the end of garment edge sewing.The surface quality of the lower tool parts directly affects the service life of the tool and the product quality of the sewing machine.At present,manual detection is the main method to identify the surface defects of the lower tool parts of sewing machine.However,the workload of manual detection is heavy,the efficiency is low,and some defects are difficult to be detected by the naked eye,and eye fatigue will also cause false detection and leakage.These problems not only limit the output and product quality of the lower tool parts,but also cause the misjudgment of the part defects,resulting in unqualified products entering the market.And they influence consumer usage.In order to improve the level and accuracy of automatic surface defect detection of lower tool parts,this paper designs a set of visual detection systems for surface defect of lower tool parts of sewing machine,and studies its detection algorithm.The research content includes the following aspects:1.Through the analysis of the causes of the defects of the lower tool parts and the location of the defects,Surface defects of lower tool parts can be roughly divided into four types: scar,collapse Angle,literal or non-literal bad,burn alive.In order to accurately detect different kinds of defects,aiming at scar,collapse Angle,and literal or non-literal defective defects,contour tracking detection method without template matching and image difference subtraction method with template matching are carried out in this paper,and the second detection method has very good detection effect.Aiming at the burning defect,this paper calculates the first and second moments of each channel through color space conversion and channel separation,and defines a burning defect judgment formula for the first time.Combined with the difference of the color distribution histogram of each channel between the qualified part and the burning part,to judge whether the part is a burning part,and observe the location and area of the defect through threshold segmentation of the V-channel image.2.In order to improve the visibility of the defect features on the surface of the lower tool parts in the image,the experiment compared four different light sources,namely,coaxial light source,Angle light source,plane light source and back light source,and investigated the image acquisition conditions from different angles.According to the image obtained by analysis,coaxial light source is selected for image acquisition.This method can ensure that the images obtained are of high quality and can reveal the defects of the lower tool parts more accurately.This approach is critical to the subsequent automated defect detection process.3.Due to the attenuation characteristic of the intensity of the light source from the center point to the periphery,the middle pixel of the acquired image has a higher threshold than the surrounding pixel,which will have an adverse effect on the subsequent image processing.Aiming at this problem,this paper for the first time carries out second-order power function,third-order power function,fourth-order power function,fifth-order power function and trigonometric function fitting for the gray average pixels of each row of the white paper image under a certain illumination intensity,and then uses the fitting results to calculate the compensation coefficient,and compensates the collected parts image through the compensation coefficient.And according to the standard deviation of the compensated experimental image and the actual situation to determine the most appropriate fitting function.4.lower tool part parts are long and thin parts with a length of more than 60 mm but a width of only 8mm.Limited by the camera resolution and the small area of some defects relative to the parts,it is difficult for a single part image to achieve the ideal detection accuracy,and multiple images need to be stitched according to the image sequence to achieve full coverage of high-precision parts information.5.A set of experimental systems for the defect detection of the lower tool of sewing machine is designed.The system mainly includes hardware part and software part.The hardware part includes computer,conveyor belt,industrial camera,sensor,light source and single chip microcomputer and other main parts.In the aspect of software,we use Visual Studio as the development environment of the system,based on MFC framework for man-machine interface design,use C++ as the programming language,and use Open CV as the function library of image processing algorithm.The system can effectively perform pixel fitting,light source compensation,image stitching and defect detection.The final test results show that the algorithm meets the requirement of no missing detection,but in some cases it cannot extract the defects completely.The maximum time spent in the simplex of defect detection is 0.597 s,which meets the requirement that the detection speed of each part in engineering detection is within 1 second and the missed detection rate is less than 1%. |
