Applications of Machine Vision in the Tobacco Industry

　　Currently, the wrapping speed of China's tobacco machinery equipment has increased from 3,000 cigarettes per minute to 10,000 cigarettes per minute, marking a significant leap from low to high speeds. As the wrapping speed has surged, gear-driven mechanisms are now operating at increasingly rapid rates, while technologies like cylindrical clamping and threaded locking have seen broader adoption. Meanwhile, with production speeds accelerating, there is an urgent need to dramatically enhance product inspection capabilities, placing higher demands on various detection technologies.

　　Visual inspection technology is one of the most promising emerging technologies in the field of precision measurement. It integrates electronics, optoelectronic detection, image processing, and computer science to bring machine vision into industrial inspection, enabling rapid 3D measurement of object dimensions or positions—whether products or components. With notable advantages such as non-contact operation, high speed, and excellent flexibility, this technology holds significant potential for applications in modern manufacturing. This article primarily explores the application of machine vision in the tobacco production process.

　　I. Applications in Removing Foreign Objects from Tobacco Strips

　　Machine vision is used on the tobacco leaf primary processing line—specifically in the re-tobacco-leaf, leaf-stem, and stem-processing stages—to remove moldy leaves, as well as foreign objects such as metal, plastic, paper, black rubber, cellophane, jute twine, thread-like materials, and non-tobacco items with varying structures.

　　The removal system consists of four main components: a feeding conveyor, a video unit, a collection bin, and an electrical control cabinet. These four components work in seamless coordination to identify, sort, and eliminate foreign objects from tobacco leaves. The system leverages advanced computer control technology, real-time digital image processing, pattern recognition techniques, artificial intelligence, and robust, reliable mechanical equipment. By employing line-array CCD technology, it captures images of single-layer tobacco leaf streams ejected horizontally at speeds ranging from 5 m/s to 7 m/s from the output end of the belt conveyor. These images are then analyzed to make precise decisions, while a rapid electromagnetic valve-based air-blowing system ensures both the speed and accuracy of the entire process—from detection to removal. Moreover, image processing is handled by four computers operating in parallel, enabling simultaneous sampling from all four cameras. Once a reference sample is established, the system requires only a single setup operation—typically taking no more than 6 minutes. This streamlined approach plays a crucial role in enhancing product quality and strengthening the company’s competitive edge in the market.

　　II. Applications in Small Package Appearance Inspection

　　Product quality is a company’s lifeblood, and cigarette packaging serves as the most direct identifier of a tobacco brand, encompassing various factors such as craftsmanship, materials, and printing. As a result, the quality of cigarette packaging plays a critical role in maximizing both the product’s intrinsic value and its perceived value to consumers. In the production process of cigarette packaging, manual spot-checks using the human eye—conducted online for quality assurance—are typically performed before introducing machine vision systems. However, this traditional inspection method often leads to numerous issues in the market, with defective products presenting a wide range of flaws. To address these challenges, machine vision-based small-package appearance inspection systems have emerged, effectively detecting many common packaging defects, such as white spots, misaligned edges, curled corners, white packages, and reversed packaging.

　　Existing applications in small-package appearance inspection have also evolved through several stages—moving from black-and-white detection to color inspection, and from standalone intelligent systems to PC-integrated solutions. With the advancement of machine vision technology, inspection capabilities across the board have significantly improved. Today, typical appearance inspection systems typically consist of: high-resolution color industrial lenses, image acquisition cards, industrial PCs, PLC control units, machine vision LED lighting sources, and specialized inspection software.

　　High-resolution industrial cameras capture images of cigarette packs and convert them into digital image signals, which are then transmitted to a dedicated image-processing system. By analyzing pixel distribution along with information such as brightness and color, the system transforms these signals into digital data. The image system subsequently performs various computations on this data to extract key features of the target object, enabling it to control the operation of rejection valves or alarm devices based on the results. This approach ensures comprehensive quality monitoring even on high-speed production lines, guaranteeing that every single product undergoes rigorous inspection—leaving no room for defective items to slip through and reach consumers, thereby protecting their interests.

　　III. Applications in Strip Packaging Appearance Inspection

　　The introduction of the strip-pack appearance inspection system ensures the quality of both the outer packaging and cellophane wrapping of strip-pack products. By conducting online inspections across multiple surfaces of the strip packs and displaying relevant data, the system accurately assesses the quality of each product. If a product fails to meet quality standards, the system flawlessly identifies and removes the defective items from the production line.

　　When selecting the hardware configuration for the strip packaging appearance inspection system, it’s crucial to comprehensively consider and design the photoelectric sensors, cameras, lighting sources, and image acquisition cards—along with their optimal installation positions—to ensure both image quality and acquisition speed. The typical dimensions of the strip package are 290 mm × 900 m × 50 mm, and the system needs to inspect four side end faces as well as one front end face. Since the cigarette packs are moving objects, a progressive-scan camera must be chosen to avoid the interlaced scanning issues like odd-even field ghosting that can occur with interlaced cameras. Additionally, based on the principle that object movement during the exposure time should not exceed 1 pixel, the system should be calibrated on-site according to the actual speed of the conveyor belt.

　　The main challenge in this application lies in simultaneously detecting all five surfaces of the cigarette stick using six synchronized cameras—balancing the massive data volume with real-time processing is particularly difficult. However, by adopting the PCBASE approach, which separates the acquisition card from the cameras, we’ve significantly enhanced the system’s detection capabilities and effectively addressed the issues at hand.

　　IV. Detection Applications Related to Cigarette Shape

　　The cigarette detection system is an essential quality-control component in cigarette packaging machines, ensuring that packs are free from issues such as missing cigarettes, misaligned filters, empty ends, or incorrectly oriented cigarettes—thus guaranteeing the highest standards of product quality before the cigarettes leave the factory. Primarily, the detection of cigarette shape ensures that the tipping paper accurately and completely covers the cigarette surface during production, with the steel stamp positioned precisely without any deviations, ultimately maintaining the consistent quality of each individual cigarette.

　　High-speed cigarette-making machines achieve production rates of over 10,000 cigarettes per minute, but to keep pace, detection systems must also match—or even surpass—this blistering speed. While many modern CCD cameras have made significant advancements, their detection capabilities still fall far short of the 10,000 inspections per minute required. As a result, identifying innovative machine vision solutions and boosting detection speeds has become critical. Fortunately, CMOS camera technology has seen remarkable progress in recent years, delivering both faster processing and enhanced detection performance—now capable of capturing images at speeds up to 1,000 frames per second. Thanks to these advancements, it’s finally possible to inspect the precise contours of individual cigarettes with unprecedented accuracy.

　　5. Applications in Robotics

　　With advancements in technology, robotics has already been widely adopted across multiple industries, and in the tobacco industry, robotic systems are increasingly used for palletizing and depalletizing tasks.

　　Robots use a variety of sensor signals—such as capacitive, inductive, and photoelectric—to detect the presence of objects and perform corresponding actions. The application of machine vision helps eliminate many potential false detections, enhancing the robot's operational capabilities while also opening up greater possibilities for advanced, intelligent applications.

　　Conclusion

　　This article explores the relevant technologies behind machine vision-based appearance inspection systems and their diverse applications in the tobacco industry. Machine vision technology has been successfully integrated into multiple production processes within the tobacco sector, delivering excellent performance while significantly reducing both labor and material costs—making its potential applications exceptionally promising. As the tobacco industry continues to prioritize quality improvement and cost reduction, the further development and deployment of visual appearance inspection systems will undoubtedly drive the industry's overall growth and advancement.