Research Of Particle Counting And Defects Inspection System Based On Visual Tracking

Computer Vision plays an important role in Artificial Intelligence, and it is also one ofmost challenge research topics in the field of science. Visual Detection is part of ComputerVision Technology, and is developed With the continuous development of Computer Visiontechnology, studying of Intelligent Control, Industrial production, and the growing in theliving standard. Visual Detection is a Interdisciplinary of Image processing, PatternRecognition, Automatic Control, Diagnostic Analysis and Artificial Intelligence etc. It iswidely used in Pharmaceutical production, Intelligent Transportation System, IndustrialProduction, Intelligent Security System and so on from1980s.Pharmaceutical Industry is one of the fastest growing industry. With the development ofPharmaceutical Industry, Pharmaceutical Production and Pharmaceutical Packagingequipment updating and improving in China, and result in the widely studying of visualdetection application in Pharmaceutical Packaging. However, Pharmaceutical detectingusing visual detection has a widely Development space in our country, concludes that bytwo reasons firstly, many Pharmaceutical manufacturers detect defection for bill by workersor test lacking bill by Photoelectric Detection. Secondly,for particle counting,The traditionalphotoelectric counting and manual counting predominate in Chinese PharmaceuticalManufacturers. Even thought, more and more scientists have paid more attention toPharmaceutical counting by visual method, the researches are mainly based on still picture,and far from on-line counting. The situation is mainly caused by two reasons: one reason isthat a special machine for particle counting and defects inspection system is need to bedevelop;the other one is that simple and efficient particle counting and defects inspectionalgorithms still needed to put forward.So,aim at efficient and accurately particle counting and defects inspecting, we putforward a visual system,and mainly deeply study the Image Processing System.The contentincludes the following five tips:Firstly, we put forward the overall design of Particle Counting and Defects InspectionVisual System,detailedly introduce the constituent parts and working principle,including each constituent part’s module designing;Secondly, on the foundation of studying recently tracking methods,present a MultipleObjects Tracking method which is based on motion interpolation, and work out a wellperformance objects tracker in VC++6.0.Thirdly, in allusion to the disadvantage s of traditional particle counting method,putforward an accuracy and efficient particle counting method based on objects tracking, andprove it by experience..Fourthly, directing at the feather of particle defects, bring up the defects inspectionmethod based on roundness which relating with border radius’s statistical variance.Fifthly, put forward the Great Probability Filter Algorithm (GPF) based nearest-neighborpixels’ gray value and compare it with traditional median filter algorithm.Experimentsshow that the GPF algorithm can denoise image and attain tiny detail in image. Concludethat it is more effective than traditional median filter.Finally,through studying the recent particles adhesion separation methods, andimprove the concave-matching method. Present the concave detecting by window-cuttingmethod.By means of amount of experiences,and the result conclude that the visual system canaccurately count particle and inspect defects,.and the accuracy rate is more than ninetypercents. The visual system are efficient enough for request, and it can process25frames persecond. the Particle Counting and Defects Inspection System is well performancevisual system.