Research On Automatic Filling Control System Based On Machine Vision

At present,the automatic filling industry is mainly using large-scale pipeline filling system,which is mainly use the sensor and PLC control system to achieve mechanical control to fill.Usually the filling device is always in the center of the bottle,where wouldn’t require how the bottle placed because of the bottle is cylindrical,so it can be filled well,but for the toxic substances filling and eccentric barrel filling,while the eccentric barrels can not locate the filling mouth by machine,so it need man to fill,and about artificial filling of toxic substances,one of it is a semi-automated,and the other is due to toxic substances filling,which is easy to affect the health of workers.In this paper,an automatic filling system based on machine vision is proposed.The system uses the knowledge of image processing to realize the application of machine vision.This article is emphasis on the design of the entire hardware measurement system and software.About the algorithm,we make the simulation and do some comparison.And we do some research about using third-party library in the embedded platform to realize image process.The main contents of this paper are as follows:(1)Introduced the hardware of the measurement system,the camera,the lens and the light source,Image processing platform,including the software of Matlab platform and embedded hardware platform and the hardware of the control platform like the PLC controller,the motor controller,the stepping motor and design PLC control programe.(2)Do some research about the preprocessing of the image.Firstly to complete gray-scale before image processing,which is convert the color image to gray-scale image,which can reduce computational complexity and it need to filter the image for the accuracy of the image measure.We do some researches about Uniform noise,pepper noise and Gaussian noise in the filtering part,and after that introduce some filter.Finally,we do some research about the edge algorithm of the image,and the effects of these operators are simulated and compared.(3)Do some researches about solutions of image measurement.Firstly,the filtering part of the image processing is improved,using two-dimensional fractional-order Kalman filter to filter the image,and using software to compare the two-dimensional fractional Kalman filter and other filtering algorithms with the effect of filtering the noisy images.Finally do some researches about classical Hough transform and random Hough transform,and summary the advantage of improved filtering algorithms,which using before random Hough transform.(4)Describes the measurement of the image,the first about calibration of the system,and analysis how to use the result of image process to transmit useful information to PLC control system,analysis the error which caused in the measurement,and come out recommendations.Make a Comparision between classical algorithms with improved algorithms,the result verified the effectiveness of the improved algorithm.After that,OpenCV library of image processing is transplanted to AM335 X embedded board,and the image measurement is realized.