Research On The Shape Matching Algorithm Based On Improved Linemod-2D And Application In The Stain Detection System On The Loading Tray

The Linemod-2D algorithm takes the gradient direction of the edge point as the matching feature,the advantage is that the matching speed is fast,but the matching accuracy is very low,which is difficult to meet the high-precision industrial application.In this paper,the shortcomings of the Linemod-2D in feature extraction,angle quantization and matching strategy are improved,at the same time,the least square adjustment method is used to correct the matching position.Based on this,a shape matching algorithm based on improved Linemod-2D is proposed(That is the algorithm in this paper).For the deficiency of feature extraction,Zernike moment is used to correct the gradient direction and position of feature points,so as to obtain more accurate feature points.And improve his feature extraction strategy,the grid method is proposed,compared with the original algorithm of Linemod-2D,the distribution of feature points is more uniform and representative.For the deficiency of angle quantization,this paper proposes to expand the quantization direction to 16,so as to improve the differentiation of quantization direction.For the insufficiency of matching strategy,this paper puts forward a strategy that combines coarse matching with fine matching.At the top(above the bottom)of the image pyramid,the improved Linemod-2D is used for fast coarse matching.At the bottom of the image pyramid,a new matching strategy is proposed for fine matching,which quantifies 360 angles.Finally,in order to further improve the matching accuracy,the least squares adjustment method is proposed to correct the matching position,obtained matching position at the subpixel level and the exact rotation angle.The experimental results show,under the influence of illumination change,noise interference and partial occlusion,compared with the original algorithm,the stability of the which is improved.In the test of image based on rigid transformation,the positioning accuracy of the algorithm is higher than shape matching algorithm of CkVisionBuilder and MIL10.0,and close to shape matching algorithm of Halcon.On the other hand,the matching efficiency of the algorithm is close to shape matching algorithm of Halcon and MIL10.0,and better than shape matching algorithm of CkVisionBuilder.The large-scale ink cartridge production line has been automated,but when the carrier transports the ink cartridge,sometimes the ink spills on the tray,forming stains.Aim at this problem,the foxConn's production line has used machine vision technology to detect ink stainson the tray.At present,in the existing stain detection system,matching algorithm of Halcon is used in the image registration,but the modeling part of the algorithm is very time-consuming,which can't meet the real-time requirements of the production line.In order to solve this problem,this paper proposes to apply the shape matching algorithm based on improved Linemod-2D to the image registration of the stain detection system.In order to verify the feasibility of the algorithm in this paper,in the experiment,the image of the ink cartridge carrier is used to test the matching accuracy,matching speed and modeling speed of the algorithm in this paper.The experimental results show that in terms of matching accuracy,the matching accuracy of the algorithm is 15.51% higher than shape matching algorithm of CkVisionBuilder,15.11% higher than shape matching algorithm of MIL10.0 and6.76% higher than shape matching algorithm of Halcon.Time consuming aspects of matching and modeling,the total time of algorithm modeling and matching is 48.06% less than shape matching algorithm of CkVisionBuilder,40.57% less than shape matching algorithm of MIL10.0and 82.06% less than shape matching algorithm of Halcon.It can be seen from this that,the algorithm has good stability,matching accuracy and real-time performance,it can be applied to the image registration of carrier plate stain detection.