Machine Vision-based Encapsulating Quality Inspection System For Glass Bottle On The Assembly Line

With the rapid development of society,people’s demand for food safety is increasing.Unqualified packaged products may appear on the automated beverage assembly line.therefore,the intelligent upgrade of quality inspection for the packaging of the beverage production line has begun to become a development trend.Traditional manual inspection has low efficiency and poor accuracy,and existing detection equipment has poor performance in certain special scenarios.To further improve enterprise production efficiency,this thesis designs a glass bottle encapsulating quality inspection system with high cost-effectiveness and strong stability,which achieves detection of abnormal bottle cap states such as uncovered,skewed,and high caps,and can stably identify liquid level states in special scenarios with bubbles or fluctuations.The main research contents of this thesis are as follows:Firstly,the overall architecture and core hardware of the system were designed.Based on the requirements analysis,the overall architecture of the system was determined.Hardware components such as industrial cameras,industrial computers,and light sources were selected.The core hardware of the system was designed,which included an ARMbased hypogynous machine,an IO signal module,and a LED controller based on dual closed-loop control of current and light intensity.The dual closed-loop LED controller used a Buck transformation circuit,and the GB32 microcontroller,utilized a cascade PID to achieve closed-loop control of the LED’s luminous intensity.This decreased the system’s cost and solved the problem of unstable luminous intensity caused by changes in the LED junction temperature.Secondly,a liquid level detection algorithm based on weighted grayscale value and region integral was designed.The algorithm addressed the difficulty of liquid level detection caused by production line fluctuations and the tendency of thick and bubbly drinks to produce bubbles.Based on the weighted grayscale value coefficient set after tuning,the algorithm introduced the concept of sliding window weighting table,conducted weighted integration on the ROI area,and combined it with the standard value of region integral to output the liquid level detection result.The test results showed that the algorithm had high accuracy under stable lighting conditions,and performed better in terms of stability and comprehensive performance in the state of fluctuating liquid levels or bubbles.Then,a bottle cap detection algorithm based on Canny and bottle body tilt correction technology was designed.The algorithm calculates the gradient and direction of the edge feature points of the bottle cap by Sobel operator,and uses the non-maximum suppression and double threshold strategy algorithm to refine and judge the edge.The detection results are outputted through angle discretization,bottle body tilt correction information,and an optimized similarity measurement formula.To address the problem of poor stability of bottle cap detection under the 180-degree imaging field of view on the production line,a360-degree image acquisition system was designed.Tests have shown that the algorithm has ideal performance.Finally,the system software was designed and the system hardware and software were tested and debugged.Software modules such as detection result management,real-time image display,historical data,lower-level machine control,product management,and production management were implemented by the MFC framework and modular design approach.The system testing showed that the system designed in this thesis has strong stability and high accuracy,and meets the requirements of practical production testing.