Research And Application Of Location Algorithm For Large TV Back Panel Based On Machine Vision

In the production process of traditional TV sets,the backplane lamp bar is fixed by manual adhesive tape.In order to reduce the labor cost and improve the automation level of the equipment,the key technology of automatic glue spray by robot instead of manual mounting is to locate accurately the visual location of the workpiece on the back panel of a TV knot,especially the positioning of the workpiece on the back panel of a large TV set under the strong interference environment in complex industrial production.Traditional visual positioning technology uses a single camera to template match certain features of the center position of the backplane,but this method has a significant error magnification effect on large TV backplates.This thesis intends to use the dual-camera fusion positioning technology,grab two corner points of the TV backplane workpiece,segment the image in a complex background,find outline points,and fit to calculate the right corner edge of the backplane.Inclination angle is calculated from the line between two right angle points to improve the accuracy of the inclination angle in the positioning data.The main research work is as follows:(1)To solve the problem of image segmentation: A combined image segmentation algorithm based on adaptive extreme median filter,improved maximum inter-class variance method and improved adaptive closed operation algorithm is presented.This method first uses adaptive extreme median filter to reduce noise interference and protect image edge information.Preliminary image segmentation is completed by the improved maximum interspecific variance method.For different types of workpieces with different edge crevice widths,an adaptive closed operation algorithm based on the target workpiece contour features is presented,which can achieve good crevice filling effect.Finally,the background interference is removed by the geometric moment of the TV back panel to achieve the image segmentation of the target workpiece in the image.(2)For the problem of contour detection: A fast algorithm for detecting image outline points based on the contour features of target workpieces is presented.Combining the contour feature of the target workpiece with the idea of traditional image contour point detection algorithm,a fast image contour point detection algorithm based on the contour feature of the target workpiece is presented by reducing the calculation dimension.(3)For the right angle positioning problem,two algorithms are proposed.One is a right-angle positioning algorithm based on multiple constraints and improved probability Hoff line detection algorithm.By setting multiple constraints,the detection rate of erroneous straight lines is reduced.By combining the advantages of random Hough transform(RHT)and probability Hough transform(PHT),an improved probability Hough line detection algorithm is presented.Two sets of outline points detected by the fast detection algorithm of image outline points based on the contour features of target workpiece are used as seed sets to reduce the blindness of selecting seed points by RHT and PHT algorithms and achieve fast and accurate positioning effect.The other is a right angle positioning algorithm for workpiece contour based on convolution and line fitting evaluation index.This method presents a new calculation method to evaluate the invalidity of the fitting effect of the least squares method for fitting a straight line when the point set approximates a parallel coordinate axis.Combining convolution with fitting effect evaluation parameters,the selection of straight line points in point set is achieved.Finally,through two straight line fitting,the straight line equation is calculated and the positioning data is obtained.This algorithm improves the real-time performance and stability while ensuring the accuracy by simplifying the calculation steps.(4)For human-computer interaction problems: The interface software of the visual system is developed.The interface software is developed with C++ programming language and MFC application framework.Store data in My Sql database;Embed the CAD drawing parser in the software to plan the optimal spray path.The communication with the lower computer is achieved by using the open source serial port library.Through the analysis of multiple groups of experiments and field production test results,the positioning algorithm in this thesis is suitable for all types and sizes of TV backplanes.Its stability is very reliable,the positioning success rate is higher than 99.9%,and its realtime performance and positioning accuracy can meet the requirements of actual industrial production.At present,the system has passed the acceptance and officially put into operation,and the annual output is expected to exceed one million units.