Research And Application Of Bottle Body Defect Detection Algorithm For Beer Empty Bottle Detection Robot

With the advent of Industrial 4.0 and the era of smart manufacturing,smart factories and smart manufacturing equipment and industries have become the common focus of all countries in the world.In Industry 4.0,the beverage intelligent production line is one of the important application scenarios.In the beverage industry,most of the production of beer uses recycled old bottles for reproduction,and strict quality inspection of empty bottles is required before canning beer.Traditional manual detection has a lot of missed inspections,and it is inefficient and costly.Using robot vision method,high detection accuracy,fault tolerance,speed,and low cost.This article mainly studies the image processing algorithm based on bottleneck defect detection unit of machine vision,aiming to develop a high-performance bottle defect detection and recognition algorithm.Before the detection of bottle defects,it is necessary to first determine the position of the bottle in the image.A good bottle positioning algorithm can greatly reduce the complexity of bottle defect detection algorithms and improve the detection speed.According to the spatial characteristics of the bottle image,this paper proposes two bottle positioning algorithms based on the center of gravity of the edge point and the extreme value of the vertical gray projection.The positioning accuracy is about 4 pixels and the average positioning time is 1ms.The experimental results show that the two algorithm can be well applied to the positioning of bottle defect detection.Due to the inconsistent light transmittance of bottle wall and the influence of LOGO text and wear on the bottle body,the gray value distribution of various parts of the bottle body is quite different,and the overall detection of the bottle body image is usually adopted.It is difficult to achieve better results.Therefore,this paper proposes a bottleneck defect detection algorithm based on the characteristics of pixel gray value distribution of bottle images.The algorithm divides the bottle image into three parts: a smooth area,a hard-banding area,and a LOGO text area.The defect recognition method based on super-pixel clustering-based defect extraction and the pixel average value of the super-pixel is adopted for the smooth area;In the wear-resistant region,a defect detection and recognition algorithm based on horizontal gradient operator is proposed.For the LOGO text region,rectangular block extraction based on improved canny edge detection operator and defect and LOGO text recognition algorithm based on convolutional neural network are proposed.The bottle defect detection algorithm designed in this paper has a detection accuracy rate of 97%,an average running time of 37 ms,a high detection accuracy,a high speed,and an international average speed of 72,000 bottles per hour,which has practical application value.