Research On Detection System Of Triode Appearance Defects Based On Machine Vision

Today's China is not only a major producer of discrete semiconductor devices,but also a major consumer country,occupying nearly half of the global semiconductor discrete device market.As one of the important components of semiconductor discrete devices,the triode is also the core component of electronic circuits.Therefore,the product quality of the triode directly affects the stability and service life of electronic products.At present,the domestic inspection methods for the appearance defects of triode is mainly manual sampling.However,this method is limited by the resolution of the human eyes in time and space and subjective factors,and the detection efficiency is low.Not only does the detection accuracy fail to meet the requirements,but also some defects are difficult or impossible to detect manually,which severely restricts the output and delivery quality of triode products.With the continuous deepening of research in the field of machine vision,the application of non-contact visual inspection technology in the field of appearance quality inspection has become increasingly mature,but there are still few related researches on defects detection for the triode placed in the barrel.Therefore,a set of appearance defects detection system for the triode in the material tube has been developed,which has good theoretical research value and market application prospects.In this paper,the triode placed in the barrel is the research object,and a machine vision-based triode appearance defects detection system is designed.The main research work is as follows:(1)The overall scheme design of the triode appearance defects detection system.Analyze the research object of this article and its appearance defects detection technical requirements,and on this basis,complete the hardware selection and layout design of the visual inspection module,as well as the overall defects detection process design;in the software structure design,use multi-threading technology to achieve Each station performs image acquisition and processing at the same time to improve the detection efficiency of the system;finally,according to the image characteristics and defects detection requirements of the triode,the image processing algorithm flow design is completed.(2)ROI(Region Of Interest)location and partition algorithm based on local feature fusion.In order to solve the problem of insufficient positioning of the triode in the material tube and interference from the groove of the material tube,local features are used for preliminary matching and positioning;for the problem of discontinuous edge of the root of the pins and other edge interference,a random sample based on RANSAC(RANdom SAmple Consensus)improved straight line detection algorithm,effectively extracting the target edge line in the image,and then fitting the straight line by the least square method to obtain the rotation angle of the triode,and correcting the rotation posture of the triode based on the affine transformation to achieve the rotation matching of the triode,And finally restore the rotation matching positioning coordinates to the original image coordinate system to complete the positioning and partitioning of each ROI of the triode.(3)Package area defects detection algorithm based on K-Means.Aiming at the problem of scratch interference on the surface of the material tube,an improved image segmentation algorithm based on K-Means is proposed,which effectively solves the interference of the material tube scratch on the defects detection;for the mixing problem in actual production,through the local difference Feature extraction and analysis to achieve mixed material detection;for surface defects in the triode package area,the standard mask image is used to extract the image of the area to be tested;for the defects in the round hole and nozzle area of the all-inclusive triode,the standard mask image is inverted After the color,the image of the area to be tested is extracted;finally,the image difference method is used to obtain the difference image,and the small area patch interference in the image is further removed by the method of morphological processing and connected domain feature analysis,and then the connection is filtered according to the defects judgment standard Domain to achieve defects detection in the package area.(4)Pins defects detection algorithm based on multi-view.Aiming at the problem that the pins surface image cannot effectively detect the pins uplift,a pins defects detection algorithm based on multi-view is proposed,and the isosceles right-angle prism is used to realize the collection of the pins end surface image.According to the detection requirements of the pins lift defect and the characteristics of the end face image,a reference positioning method based on Canny is proposed.The edge detected by Canny is traversed and searched and statistically determined to extract the edge of the barrel,and then the least square method is used to obtain the barrel.Fit the straight line of the edge and use it as the reference for detecting the position of the pins end face;then search and locate the center coordinates of the pins end face based on the connected domain,and calculate the horizontal distance between the center of the pins end face and the edge reference.Perform statistical analysis on the relative position between the two to achieve a more robust pins lift detection.Aiming at the problem that the standard mask cannot be used due to the uncertain contour of the thin pins,an adaptive mask generation method is proposed,which can realize the generation of the thin pins mask image with different contour shapes,and through the multi-feature combination of pins bending angle,pins length and defects size,to realize the detection of pins surface defects.To sum up,according to the technical requirements of the triode appearance defects detection,this paper designs the triode visual inspection module and the overall defects detection process,studies the triode ROI positioning and partition algorithm,and the triode package area and pins appearance defects detection algorithm,and finally based on Visual Studio 2015 software platform and modular programming ideas,developed a machine visionbased triode appearance defects detection system software,including user login module,detection parameter configuration module,online detection module,data management module and Mapping display module.After a certain scale of field production test verification,the machine vision-based triode appearance defects detection system designed and developed in this paper can complete the appearance defects detection of a triode within 150 ms,which meets the working rhythm of actual production requirements;at the same time,the average over-detection rate and miss-detection rate of the defects detection system are 1.14% and 0.08%respectively,and the defects detection accuracy rate is above 97.87%,which can meet the accuracy and stability of actual production requirements.