Research On System Architecture And Light Source Optimization Strategy For Paper Defect Diagnosis System Based On Machine Vision

Paper defects are product defects on the surface of paper that do not meet the quality standards,such as black spots,holes,pulp spots,etc.For some special purpose paper occasions,such as aerospace recording paper,industrial capacitor production paper,etc.,the emergence of paper diseases will cause great economic losses,such as aerospace safety and industrial production accidents.In order to reduce the impact of paper disease on people's production and life,it is particularly important to accurately find paper disease in the process of paper production and remove the paper disease area in time.The paper disease diagnosis system based on machine vision has the characteristics of high diagnosis accuracy and fast response.It can quickly and accurately realize the on-line diagnosis of paper disease.At present,the research on paper disease diagnosis system based on machine vision at home and abroad mostly focuses on image segmentation and image recognition after paper disease image acquisition,and there is less research on system architecture and light source optimization strategy.Previous research shows that the unbalanced hardware configuration of the system leads to low diagnosis speed,poor uniformity of light source leads to poor gray uniformity of the image collected by the system,which will cause great difficulty in rapid and accurate judgment and recognition of paper disease.Therefore,this thesis focuses on the hardware architecture design,light source module design and paper disease determination algorithm of the paper disease diagnosis system.The main work and contributions of this thesis can be summarized as follows:(1)Design and research of economic and practical paper disease diagnosis systemBy analyzing the width and detection accuracy,the total camera resolution is calculated;By analyzing the speed and detection accuracy of the paper machine,the required camera acquisition frequency is calculated;According to the horizontal resolution and acquisition frequency of the camera,the amount of data generated by the camera per unit time is calculated,and the design and research of an economic and practical paper disease diagnosis system is carried out.Taking the amount of data generated by the camera in unit time as the main line,the processing capacity of the hardware equipment corresponding to the three modules of paper image data transmission,storage and preprocessing is compared and analyzed.On the basis of meeting the speed and width of the paper machine,the appropriate FPGA model is selected according to the data processing cost in unit time;A distributed paper defect determination scheme is proposed,and an FPGA image analysis card is configured for each camera to determine the existence of paper defects in real time;For the paper disease found by the paper disease determination module,the corresponding host computer is used to complete the identification of paper disease types.For the purpose of configuration balance,the architecture of paper disease diagnosis system based on machine vision is given.According to the data processing requirements,the balanced hardware configuration is selected to reduce the system cost on the premise of ensuring the timely processing of data.(2)Research on optimization strategy of transverse light source based on near-field uniform illuminationIn the experiment,it is found that when the light beads on the light source module are unevenly distributed,there is an obvious gray unevenness in the transverse direction of the image collected by the paper disease diagnosis system,which makes the threshold method with the highest efficiency of paper disease determination unable to be applied.In this thesis,based on the study of the irradiance theory in geometric optics,combined with Spyro's law,the optimal distance function between the linearly arranged led beads in the light source module is obtained in the near light field.At the same time,the light field with the highest uniformity produced by the linearly arranged beads is determined according to the formula,and the diagnosed paper is placed on the near light field,the illumination uniformity of the light on the diagnosed paper irradiated by the light source is realized.Combined with the related concepts of secondary light source,the fact that the camera passively receives light to generate images is determined.Based on the principle of diffuse light decomposition,the power optimization function of light beads at different positions in the light source array is determined by experimental method.The paper as the secondary light source can decompose the light to the camera lens to achieve uniformity.Finally,the image with uniform background gray level in the wide image acquisition system is obtained,which lays a foundation for the simplification of paper disease determination algorithm and paper disease recognition algorithm and the improvement of program execution efficiency.(3)Research on optimization strategy of LED driving power supply based on stroboscopic imaging principleIn the experiment of high-speed and wide width paper disease diagnosis system,it is found that under the state of high precision and high speed,in order to ensure that the collected image does not produce distortion,the paper image acquisition operation needs to be completed by increasing the camera acquisition frequency.At this time,the single acquisition time of the image acquisition module reaches the microsecond level.In this case,the image acquisition module will be more and more sensitive to the fluctuation of the light source in unit time.Under the irradiation of ordinary linear light source,the image with transverse ripple will be collected,and the gray uniformity of the image background is poor,which is not conducive to the later image analysis operation.Through the reverse analysis of Shannon's sampling theorem,this thesis determines that the reason for the transverse ripple is that the frequency of the light source is less than that of the camera;By analyzing the physical characteristics that led beads do not produce stroboscopic,it is determined that the main reason for stroboscopic is the instability of LED driving power supply;Combined with the principle of stroboscopic imaging,the LED driving power supply circuit based on the principle of stroboscopic imaging is designed,so that the camera can obtain equal luminous flux in each acquisition cycle,effectively avoid the image gray change caused by current fluctuation,improve the imaging effect of the camera during high-frequency acquisition,effectively reduce the algorithm complexity of the diagnostic program and improve the execution efficiency of the program.(4)Research on adaptive threshold paper defect determination method based on dynamic optimizationThe image effect collected by the paper disease diagnosis system based on machine vision is easy to be affected by external light,resulting in large changes in the gray value of the image collected by the paper disease diagnosis system,so that the phenomenon of misjudgment often occurs in the application of threshold method.The universality and stability of the system are poor,which is not conducive to the promotion of the system.To solve this problem,this thesis proposes an adaptive threshold paper disease determination method based on dynamic optimization.The image gray data in different time periods are obtained by experimental method.The image gray data are classified according to the time period of the obtained data,and the maximum average value,minimum average value,overall average value and standard deviation of gray in the corresponding time period are obtained.The chart analysis method is used to compare and analyze various image gray data,summarize the image gray change law,and program and verify it with Visual studio 2015.Experimental results show that this method can better adapt to the changes of external light and improve the robustness of paper disease determination system.To sum up,this thesis adopts the research method of combining theoretical and experimental research to better solve the technical problems of poor imaging effect,high complexity of detection algorithm and low efficiency of system execution in the paper disease diagnosis system based on machine vision.The experimental results show that the solution proposed in this thesis has important guiding significance to improve the diagnosis accuracy,diagnosis speed and application width of the system.In addition,the research results in the design of light source module can provide theoretical support and technical guidance for the accurate and rapid acquisition of image information of high-frequency wide web paper machine.