Visual Inspection Technology for PET Bottle Appearance Defects: From Principles to Industrial Applications

1. Common Appearance Defects in PET Bottles and Their Impact on Product Quality

As a primary packaging form for beverages, food, and other fast-moving consumer goods, the appearance quality of PET bottles directly affects consumers' first impression of the product and is crucial to the safety and shelf life of the contents. During production, PET bottles undergo multiple processes, including blow molding, filling, labeling/applying, and coding. Each process can introduce specific appearance defects. In the blow molding stage, when high-temperature gas is blown into the preform and pressed into shape by a mold, defects such as bottle deformation and crooked necks are prone to occur. In the filling stage, common problems include missed filling, insufficient filling, and loose or crooked caps (including high caps and crooked caps). Labeling and applying processes may result in misaligned or missing labels, while coding processes may lead to missed or incorrect coding due to equipment malfunctions.

These defects range from minor impacts on the product's appearance to serious consequences that directly alter the quality of the contents. Especially for oxygen-sensitive products such as tea beverages and dairy products, improperly sealed bottle caps can lead to product spoilage, while incorrect labeling or unclear coding can cause market supervision issues. Therefore, the detection of appearance defects in PET bottles is not only related to product appearance but is also a key link in quality control.

Table: Main Defect Types of PET Bottles and Their Impact

Defect Category	Specific Defect Example	Impact on Product Quality
Bottle Cap Defects	High cap, crooked cap, broken cap, no cap	Contents spoilage, leakage and contamination
Label Defects	No label, high or low label, perforated label, crooked label	Incomplete information, impact on brand image
Coding Defects	Missing coding, blurry, incorrect date	Difficulty in product traceability, market supervision risks
Bottle Body Defects	Deformation, crooked neck, leakage	Packaging function failure, loss of contents

2. Technical Principles and System Composition of the Visual Inspection System

The fully automated visual inspection system simulates human visual function, using image acquisition equipment to acquire images of the PET bottle's appearance. Then, advanced image processing algorithms analyze the image quality to ultimately achieve defect identification and classification. A complete visual inspection system typically consists of three core parts: an image acquisition module, an image processing module, and a result execution module.

The image acquisition module, the "eyes" of the system, mainly consists of an industrial camera, optical lenses, and a dedicated light source. Given the highly reflective nature of PET bottle surfaces, a specially designed lighting system is typically required to highlight target features. For example, the inspection equipment developed by Jinan Maotong employs a "combined front and backlighting shooting method" and a "telecentric optical configuration," effectively highlighting labels and their defect features, ensuring accuracy and high precision in inspection. For the inspection of transparent bottles, specialized optical sensors (such as the 5-series and 3C-series) provided by companies like Leuze can effectively address the challenges of inspecting transparent materials.

The image processing module, the "brain" of the system, undertakes the most critical defect identification task. This module first preprocesses the acquired images, including grayscale transformation, threshold segmentation, and binarization, to enhance useful defect information in the images. Subsequently, through algorithms such as edge detection, feature extraction, and pattern recognition, the system can accurately determine whether the product is qualified. For example, in bottle cap inspection, the system calculates the maximum distance from the top edge of the cap to the baseline and determines whether the cap seal is qualified by using the angles between the left and right edge lines and the baseline and top edge lines.

The result execution module acts as the system's "hand," responsible for translating processing results into actual actions. When the system detects a defective product, it immediately sends a signal to the rejection device to accurately remove the non-conforming item. To improve rejection accuracy, the system typically incorporates a speed encoder to automatically track the conveyor belt speed, achieving synchronous rejection. Modern vision inspection systems also possess rich data statistical functions, capable of displaying the quantity and proportion of non-conforming products in real time on a touchscreen and enabling remote monitoring and storage of production data via an Ethernet interface.

3. Technical Approaches of Different Types of Vision Inspection Systems

PET bottle appearance defect vision inspection systems can be broadly categorized into two technical approaches based on their core processing technologies: embedded systems based on programmable chip-on-a-chip (SOPC) and industrial integrated systems. Each approach has its advantages and is suitable for different production scenarios and needs.

3.1 Embedded Vision Inspection Systems Based on SOPC Technology

Inspection systems based on programmable system-on-a-chip (SOPC) technology utilize FPGAs (Field-Programmable Gate Arrays) as their hardware core, fully leveraging the parallel processing capabilities of programmable logic devices. These systems are typically based on Altera or Xilinx FPGA chips, with soft-core processors such as Nios II configured as the control core. The system achieves complete image processing functionality by connecting peripheral modules such as image acquisition interfaces, storage controllers, and VGA controllers to the Avalon bus.

A significant advantage of this type of system lies in its high-speed processing capability. Because FPGAs enable hardware-level parallel computing, image processing speeds are far superior to systems based on general-purpose processors. Experimental data shows that an SOPC-based system processes a single bottle image in approximately 90 milliseconds, translating to a detection speed of over 10 bottles per second and an accuracy exceeding 99%. Simultaneously, this system offers high flexibility, allowing users to customize image processing algorithms and detection standards as needed, such as implementing specific image detection judgment criteria through C++ programming.

However, the SOPC technology route also requires companies to possess strong hardware design and algorithm development capabilities, making it more suitable for large beverage manufacturers or system integrators with extreme requirements for detection speed.

3.2 Industrial Integrated Vision Inspection Systems Industrial integrated systems are built using mature industrial vision components, such as Siemens vision processors combined with PLCs (Programmable Logic Controllers) and touchscreens to form complete inspection systems. These systems are typically developed as overall solutions by specialized automation companies (such as Hangzhou Huafeng Automation Systems Co., Ltd.), providing complete functionality from image acquisition to defect removal.

The advantages of industrial integrated systems lie in their high stability and rapid implementation. The system utilizes all stainless steel materials, meeting the hygiene standards of the food and beverage industry; its independent mechanical structure allows for easy installation at any location on the production line; and its X-Y-Z adjustable mechanism provides greater operational flexibility. For example, Siemens vision systems offer both digital signal interfaces and industrial network interfaces, allowing users to select the appropriate configuration based on their actual working conditions, which is both convenient and cost-effective.

These systems also excel in inspection speed, achieving speeds of over 1500 bottles per minute in practical applications. Furthermore, the system possesses powerful data communication capabilities, communicating with computers via Ethernet interfaces to enable online monitoring, program modification, and production data recording, providing a data foundation for intelligent manufacturing.

Table: Comparison of Visual Inspection Systems with Different Technical Approaches

4. Practical Application Cases and Performance Analysis

The practical application effect of visual inspection technology in the detection of appearance defects in PET bottles is remarkable. Different manufacturers' dedicated equipment has its own characteristics in terms of detection speed, accuracy, and adaptability. The following is a performance analysis of several typical application scenarios.

In terms of bottle cap inspection, Jinan Maotong's PET bottle all-around inspection machine can detect various defects such as no cap, high cap, crooked cap, broken bridge, broken ring, mixed cap, and color difference cap. The inspection speed is up to 36,000 containers per hour, equivalent to 600 bottles per minute. This equipment covers standard 28-cap, 38-cap, double-layer cap, and other non-standard cap types, almost covering the mainstream PET bottle cap types on the market. The installation location is typically after the capping machine, enabling timely detection of defective products and preventing defects from flowing into subsequent stages.

Label inspection is another key application. Jinan Maotong's 360° omnidirectional combined label inspection machine adopts a multi-station architecture, observing from six different angles to achieve 360° blind-spot-free inspection. The equipment uses a "synchronous-asynchronous triggering mode" and a patented splicing detection algorithm to accurately identify defects such as missing labels, joint labels, uneven labels, perforated labels, inverted labels, cut labels, cracked labels, wrinkled labels, and horizontally offset labels. This equipment boasts an impressive inspection speed of up to 48,000 containers per hour, meeting the demands of high-speed production lines.

For filling level and sealing inspection, X-ray level detectors and squeeze leak detectors provide effective solutions. X-ray detectors utilize the principle that different substances absorb X-rays differently, enabling them to penetrate various transparent and non-transparent containers to detect liquid levels, with a speed of up to 60,000 containers per hour. The squeeze leak detection machine uses multiple liquid level sensors and high-precision pressure sensors to jointly determine the capacity and sealing performance of the container by analyzing the liquid level and internal pressure feedback values under different squeezing intensities, enabling it to sensitively detect even minute leaks.

It is worth noting that these detection devices not only can reject defective products in real time, but also possess rich data statistical functions. For example, the system developed by Hangzhou Huafeng can statistically analyze the number and proportion of defective products passing through each unit of product, the number and proportion of defective products passing through each unit of product within a unit of time, and the total production volume and number of defective products for each shift and month. This data is displayed in real time on a touch screen and can be connected to a printer for output, providing a basis for production quality management decisions.

5. Technical Challenges and Development Trends

Although visual inspection technology has achieved significant results in the detection of appearance defects in PET bottles, it still faces some technical challenges. At the same time, to meet the increasing demands for quality control, this technology is constantly evolving.

5.1 Current Technical Challenges

Detecting transparent materials is one of the main challenges faced by visual inspection systems. PET bottles themselves have transparent or semi-transparent properties, which easily generate reflections and refractions, interfering with image acquisition quality. To address this challenge, sensor manufacturers have developed dedicated solutions. For example, Leuze's 5-series sensors can detect translucent bottles and transparent films; the 3C series is specifically designed for detecting transparent objects and features tracking capabilities. However, stable detection under complex lighting conditions still requires sophisticated optical design.

The real-time requirements of high-speed production are another major challenge. Modern beverage production lines are constantly increasing in speed, with hundreds or even thousands of bottles processed per minute becoming commonplace. This places extremely high demands on image acquisition and processing speeds. FPGA-based parallel processing architectures are an effective way to solve this challenge, significantly improving processing speed through hardware-level parallel computing.

Furthermore, the identification of complex defect types requires more advanced algorithms. For example, accurate identification of defects such as label wrinkles and slight misalignment requires algorithms with strong anti-interference capabilities and pattern recognition abilities. Combining traditional image processing algorithms with modern artificial intelligence technology may be the future direction.

5.2 Future Development Trends

The future of PET bottle visual inspection technology will move towards higher precision, faster speed, and greater intelligence. With the improvement of camera resolution and the enhancement of processor computing power, detection accuracy will break through from the current micrometer level to the nanometer level, enabling the identification of even smaller defects. Simultaneously, detection speed will continue to increase with the improvement of processor performance, meeting the needs of ultra-high-speed production lines.

The deep integration of artificial intelligence and machine learning is another important trend. Through deep learning algorithms, the system can autonomously learn defect features from a large number of samples, reducing the complexity of algorithm debugging and improving recognition accuracy, especially for complex defects that are difficult to describe using traditional rules. Furthermore, quality prediction functions based on big data analysis will also become possible, realizing the transformation from "post-event detection" to "pre-event prevention."

System integration and multi-functionality are also important development directions. Modern vision inspection systems are no longer limited to single functions but are developing towards integrating multiple detection functions into one unit. For example, a single device may simultaneously integrate bottle cap, label, liquid level, and inkjet coding detection functions, reducing the equipment footprint and improving overall inspection efficiency. At the same time, the deep integration of the system with other equipment on the production line (such as industrial robots, AGVs, etc.) will form a closed-loop intelligent manufacturing system of "detection-judgment-sorting-optimization."

With the deepening advancement of Industry 4.0 and intelligent manufacturing, visual inspection systems will no longer be isolated quality control units, but will be deeply integrated into the entire production management system, achieving data sharing and process optimization, and providing comprehensive quality assurance for PET bottle production.

Conclusion: 

Fully automated visual inspection technology for PET bottle appearance defects has become an indispensable and important means of quality control in the modern beverage industry. From FPGA-based custom systems to industrial integrated solutions, various inspection equipment plays a key role in improving product quality, reducing production costs, and increasing production efficiency. Facing challenges such as transparent material inspection, high-speed production, and complex defect identification, this technology continues to innovate and evolve towards higher precision, faster speed, and greater intelligence. With the deep integration of advanced technologies such as artificial intelligence and big data, the application prospects of visual inspection technology in the PET bottle packaging field will be even broader, providing manufacturers with more comprehensive and reliable quality assurance solutions.