The development of machine vision predates deep learning by more than 40 years, promoting industrial modernization. It is mainly applied in two directions: detection and machine vision. Therefore, it plays a huge role in the quality inspection of glass bottles. It replaces human eyes to detect products, with higher accuracy and faster speed. Therefore, glass bottle manufacturers at home and abroad widely use quality inspection equipment based on machine vision.
The development of glass bottle quality testing equipment in foreign countries predates that in China, one of the reasons is that beer, which ranks third in sales in the beverage industry, was first produced abroad. The increase in its sales has driven the development of the beer bottle industry. In order to promote industrial modernization, foreign R&D teams have begun to devote themselves to the research and development of modern inspection machines that can replace manual labor and improve production efficiency. In the 19th century, foreign countries began to develop empty bottle quality inspection machines based on machine vision technology, and corresponding inspection systems were established (mainly including image acquisition, image processing and recognition, power removal devices, and data statistics). Companies with mature technology include NI (National Instruments) in the United States (Liu Lin, 2007), which used visual inspection technology to combine machine vision, control systems, and LabView virtual instrument software to achieve a complete system. LabView software is NI’s core design platform, and this development environment is widely used in industrial data acquisition and instrument control; SGCC International in France, many domestic manufacturers are using its M1 fully automatic multifunctional glass bottle and can online inspection machine; Siemens AG (Zhu Mingzhu et al., 2014) has successfully developed the intelligent industrial vision system SIMATICV710, which has the advantages of integration and distributed processing. It integrates image acquisition equipment, processing equipment, and bus interfaces together, providing networking functions and external interfaces (Ni Zheng et al., 2004); InnoLas, a German company, has advanced camera equipment that can capture images clearly and quickly, which is beneficial for extracting defect features of glass bottles in later stages; AGR, a Japanese company, has fully functional equipment mainly used in the production of glass bottles and cans and beverage filling; Industrial Power Machinery Co., Ltd. in the United States has launched a machine for comprehensive inspection of glass bottles. Based on reflective optics, high-resolution photography technology and automatic zoom technology are applied to achieve high-precision inspection, with a detection speed of up to 800 bottles/min (Man Dehu, 2013); In addition, there is the 186 type simulated impact testing machine from the United States, which will be crushed if it fails the pressure detection. This method is relatively simple and has a fast detection speed, but its safety is relatively low; The 237/238 bottle body rotating crack detection machine from a Swiss company uses the principle of light reflection and refraction to inspect various cracks on glass bottles; The SE218 pre selector produced in Switzerland is used to detect appearance defects such as bottle height, diameter, bottle mouth deviation, bottle body skewness, and sinking. In addition, there are companies such as Germany’s Haifu and the UK’s PLM-Red_fearm (Xia Xianhua et al., 2013).
At present, domestic glass bottle manufacturers are involved in more and more products, such as beer bottles, white wine bottles, soy sauce bottles, etc., of which beer bottles were first put into production. At that time, there was no mature quality inspection equipment in China, and foreign equipment was expensive. So at first, each factory used manpower to conduct quality inspection, and then gradually purchased visual inspection machines. However, most of the machines are imported from abroad by Chinese agents. One machine is about 900000 yuan. For factories with a daily output of more than 1 million bottles, at least 8 inspection machines are needed. There are also companies in China developing glass bottle quality inspection equipment. The technically mature equipment mainly includes the Saturn bottle inspection machine developed and produced by Beijing Saiteng Power Co., Ltd. and the DS empty bottle inspection machine jointly developed by Guangzhou Dayuan and Beijing Sitong Electric using Japanese vision systems. These devices all use machine vision, pattern recognition and other technologies. However, with the increase of production volume and types, the original technology gradually cannot meet the demand. One is speed, which can only be solved by increasing the number of inspection machines and widening the retention area, but the cost and risk also increase; One issue is that its defect recognition is not complete enough, classification is inaccurate, and so on. It is understood that deep learning has not yet been involved in the quality inspection technology of glass bottles at home and abroad, but many other industrial production defect inspections have already begun research in this area.