Visual Inspection Technology for Can Bottom Coding: Achieving Zero Defects on High-Speed Production Lines

Inspecting 72,000 cans per hour with an accuracy rate of up to 99.99%—How Machine Vision Technology Revolutionizes Quality Control in the Food and Beverage Industry

In today's rapidly developing food and beverage industry, can production lines have reached speeds of 72,000 cans per hour, equivalent to an astonishing 20 cans per second. In this high-speed production environment, quality inspection of can bottom coding has become a critical step in ensuring product quality. As a carrier of product information, coding is directly related to product traceability and quality control.

Traditional visual inspection is no longer able to meet the demands of high-speed production. The introduction of machine vision technology has revolutionized this situation, enabling efficient and accurate automated inspection.

1. Causes and Challenges of Can Coding Quality Defects

In the can beverage production process, various factors can lead to coding quality defects. A water blower is typically installed before the inkjet printer to remove water droplets from the can bottom coding area. Improper adjustment or displacement of this device can result in incomplete removal of water droplets from the can bottom, resulting in blurred or missing codes. The printer itself may also malfunction. Clogged ink nozzles are a common problem, often caused by extended use or neglected cleaning. If not detected promptly, this can lead to production accidents such as missed or missing codes. Furthermore, displacement of the printer's trigger sensor or printhead can cause problems such as missing characters, missing codes, missing codes, or incorrectly positioned codes.

The concave bottom of metal cans presents additional challenges for visual inspection. Compared to conventional transparent materials, the reflective nature of metal surfaces makes imaging more difficult, requiring specialized lighting solutions to achieve stable and clear images.

2. Design Goals and Key Technical Parameters of the Inspection System

To meet the demands of high-speed production lines, the can bottom code inspection system must meet a series of stringent technical specifications.

High-speed real-time performance is the most fundamental requirement. The system must complete image acquisition, analysis, and judgment instantly as the can passes through the inspection area at high speed, requiring processing speeds in the millisecond range.

Equally critical is rejection accuracy. The system must accurately distinguish between qualified and unqualified products, rejecting only defective ones to avoid losses caused by false rejections. System scalability is also crucial. Flexible I/O interfaces facilitate integration with various sensors or actuators to meet diverse on-site measurement and control needs.

Data statistics and communication capabilities are key features of modern inspection systems. The system should be able to record, compile, and display inspection data in real time, while also enabling data exchange and remote transmission via Ethernet or serial interfaces.

3. System Architecture and Core Hardware Components

A complete can bottom inkjet inspection system consists of three main components: a light source and vision processing system, an electrical control and human-machine interface system, and a defective rejection device.

Light Source Configuration

Due to the concave surface and reflective metal properties of aluminum can bottoms, the system typically uses a spherical integrating light source. This light source features a hemispherical inner surface that evenly reflects light emitted from the bottom 360 degrees, ensuring uniform illumination across the entire image acquisition area and significantly improving image quality and stability.

Smart Camera Selection

The smart camera is the core of the system. The Cognex In-Sight Micro1400 smart camera, with its compact size (only 30mm x 30mm x 60mm) and powerful functionality, is particularly well-suited for high-speed assembly line applications. Embedded with sophisticated machine vision algorithms, it performs functions such as presence/absence detection, surface defect inspection, dimensional measurement, and OCR recognition. This requires minimal user programming, significantly accelerating system development.

Processing Unit Configuration

Systems typically utilize a combination of an industrial computer and a PLC. For example, the Advantech 1261H touchscreen industrial computer paired with a Siemens S7-200 PLC (with a 224-bit CPU) ensures speed and stability while maintaining cost-effectiveness.

4. Image Processing and Character Recognition Algorithm Evolution

With technological advancements, character recognition algorithms for can bottom inkjet printing have undergone significant evolution, from traditional methods to advanced deep learning techniques.

Character Area Positioning

Character area positioning is the first step in the recognition process. On actual production lines, cans are prone to rotation during the printing process, making it impossible to set a fixed ROI. The MSER method (Most Stable Extrema Region) effectively achieves coarse localization of character regions by binarizing the image and applying different grayscale thresholds. This method, combined with morphological dilation and area methods, achieves fine localization, using the minimum enclosing rectangle to determine the character region and rotational orientation.

Character Segmentation Technology

Due to the characteristics of dot-matrix inkjet characters, traditional projection segmentation techniques often struggle to accurately segment them. The dot-matrix character segmentation algorithm employs the waveform dilation method, improving upon projection segmentation. By setting segmentation thresholds and handling overlapping characters, it effectively addresses the dot-matrix character segmentation challenge.

Character Recognition

Character recognition algorithms have evolved from traditional feature extraction to deep learning. Convolutional neural networks (CNNs) autonomously extract high-quality inkjet features through machine learning, avoiding the complex manual feature extraction process. They offer excellent fault tolerance and parallel processing capabilities, significantly improving recognition accuracy.

5. System Workflow and Quality Control

The workflow of the can bottom inkjet inspection system is a highly collaborative process.

As cans pass through the imaging system, a metal proximity switch triggers a strobe light source and an industrial smart camera to capture images of the can bottom in high-speed motion. The intelligent camera then analyzes and processes the image, determining whether the inkjet printing quality meets standards and transmitting the results to the electrical control system.

For any defective products detected, the system activates a rejection mechanism to automatically remove them from the production line. This fully automated process requires no human intervention, ensuring continuous and efficient production line operation.

The system also records and compiles various inspection data in real time, providing data support for production quality management. Operators can monitor system operating status through the human-machine interface and adjust parameters promptly to ensure optimal equipment efficiency.

6. Application Results and Future Outlook

Practical applications have demonstrated that the machine vision-based can bottom inkjet printing inspection system has significant advantages.

Compared to manual inspection, machine vision systems are not only faster but also significantly more accurate, achieving 99.99% recognition accuracy. The non-contact inspection method avoids contamination during the inspection process and reduces labor costs.

With the continuous improvement of automation in the food and beverage industry and the continued increase in labor costs, the value of promoting and applying machine vision inspection systems is becoming increasingly prominent. This technology also has positive implications for breaking the monopoly of foreign equipment. In the future, with the further development of deep learning algorithms and improvements in hardware processing capabilities, can bottom coding inspection systems will develop towards higher speeds, higher accuracy, and greater adaptability, providing even more powerful technical support for quality control in the food and beverage industry.

In the future, with the continuous advancement of artificial intelligence technology, can bottom coding inspection systems will become even more intelligent and adaptive. The introduction of deep learning algorithms will enable the system to handle more complex coding defect types, while the application of 5G technology will enable remote monitoring and maintenance, further reducing operational costs.

For food and beverage manufacturers, investing in advanced visual inspection systems is not only a necessary means of quality control but also a strategic choice for enhancing brand value and market competitiveness. In the pursuit of zero defects, machine vision technology is becoming an indispensable tool.