Research On Beer Bottle Bottom Defect Recognition Algorithm For Empty Bottle Detection Robot

In the food and beverage industry,as food and beverage before filling,defect detection should be carried out on the empty bottles.And the empty bottle inspection robot based on machine vision technology compared with traditional manual testing,has advantages of high efficiency,low cost.So its research in the food and beverage industry is valuable.In this paper based on machine vision technology in the empty bottle inspection robot bottle positioning and defect detection algorithm for research.Including the bottle image acquisition scheme design and improving the method of the bottom control system software design optimization,and the research on bottle positioning algorithm and defect detection identification algorithm.At present,the bottle bottom localization algorithm remains positioning center inaccurate,consuming long time.The bottom of the bottle defect detection accuracy rate also needs to improve.In the analysis process of the bottle bottom positioning method of the empty bottle testing robot,this paper first adopted the traditional center positioning methods,such as the center of gravity method,least square method,Hough transform method,three-point circle fitting respectively fixing the bottle of the circle,analyzes the advantages and disadvantages of several kinds of positioning methods,summarizes the bottle image positioning error of the cause.According to the geometrical characteristics of the bottle slip lines,advance an modified method based on slip lines position of multiple random circle fitting location method.First in the bottom of the bottle image preprocessing,try to eliminate noise in image,then use the gravity method is scheduled to obtain initial center of the circle,the radial scan to extract real antiskid lines edge points,it will be the input edges for the circle fitting method.At last,with several times random circle fitting method for bottle center positioning,determine the final position of the circle.This method is effective to reduce the influence of noise points for positioning.Compared with several other positioning methods,this method obtains a good positioning result on the precision and speed.Through the specific studies of the beer bottle bottom image defect types and characteristics,put forward a way that the bottle image is divided into non-slip regions and central region of two component area of defect detection.For the center area of the bottle bottom,using the improved method of identification of the center area of the minimum moment center.For the anti-skid area of the bottle bottom,the identification method of the bottom defect of the anti-slip zone based on the region is adopted.And uses the combined with radial basis kernel function of support vector machine(SVM)to classify bottle defect features.Finally,designed a n empty bottle detection software system,to verify the algorithm of this paper,the experimental results show that the bottom of the bottle positioning error is less than six pixels(in this paper,the experimental object size is 648 * 483 pixels),the bottle detection time less than 100 ms,The positioning accuracy and detection speed are better than that of the previous bottle bottom localization algorithm.The defect detection accuracy is 92.7%,it has a real education practical value.