Research On Quality Detection Algorithm Of Beverage Bottle Based On Robot Vision

Due to the fact that PET bottles have the technical advantages of light weight and good preservation performance,and they lay great emphasis on heat resistance and pressure resistance,therefore,PET bottles have become the mainstream of today's drinking packaging.Besides,their quality has also attracted more attention due to their wider application.However,the human eyes have some unfavorable factors,such as low efficiency of detection,high rate of false detection,and bad working conditions.Therefore,the development of a high-speed and high-precision machine vision-based image processing algorithm can not only improve the production efficiency of the beverage production line,but also liberate the workforce to a large extent.This paper presents a multi-function detection algorithm for PET bottles in the basis of machine vision.In the light of presented multi-function detection algorithm,a set of inspection robots have been devised and implemented,which can inspect PET bottle packaging quality with high-speed and high-precision.The analysis of the functions of the system mainly includes the following four functions:pre-entry detection,image acquisition,defect detection,precise rejection,and visual operation interface.As for image acquisition,a set of imaging system was designed to achieve multi-angle acquisition of PET bottle images.With regard to the visual operation interface,a human-computer interaction system was developed by using MFC so as to realize the control of the entire system.In terms of algorithm design,PET bottles were analyzed to identify the unqualified PET bottles into three labels,namely bottle cap quality,anti-theft ring quality,and liquid level quality.1)The unqualified bottle caps have the following three characteristics:high caps,crooked caps and bottle without caps.This paper makes use of the algorithm of the combination of contour linear equation and support ring linear equation of the upper cap to make defect judgment;2)The unqualified anti-theft rings have the characteristic of dropping off the anti-theft ring,this paper make use of the slope of the support ring to make angle compensation algorithm against the image rotation so as to determine the specific condition of the anti-theft ring;3)The unqualified liquid level is mainly due to the fact that the liquid level is too high or too low.This paper utilizes the connected domain-based algorithm for liquid level position so as to determine whether the liquid level is qualified or not.The three algorithms are not mutually independent.The post-sequence algorithm can take advantage of the previous relevant detection parameters to reduce detection time.Moreover,the algorithm relies on the characteristics of the supporting ring relative to the bottle body,which can eliminate the impacts of the bottle body flicker upon the detection result.According to the imaging characteristics,algorithms were designed.Moreover,algorithms are interlocking,which can not only improve the detection accuracy of the algorithm but also shorten the detection time,which achieved a high-speed and high-precision detection targets.The algorithm proposed in this paper can be applied to the inspection system of PET bottle packaging quality in industrial production lines.After testing on the experimental platform,the detection accuracy of the algorithm can reach over 99%,and the detection time of a single bottle is about 43.1ms,which is higher than the current detection rate of 42,000 bottles/h of PET bottle inspectors at home and abroad,therefore,the algorithm meets the production requirements of high-speed and high-precision automation,which has strong practicality.