Chapter 57: Machine Vision - Defined |
Summary |
Machine vision is the technology that allows automated systems to 'see' and interpret the world using cameras, processors, and algorithms. In the context of barcoding, it signifies a fundamental shift from traditional laser scanners to intelligent cameras that can read any symbology under any condition---regardless of orientation, distortion, or challenging lighting. This chapter explores the technical definition of machine vision, contrasts it with computer vision, and delves into the practical, real-world applications across logistics, manufacturing, healthcare, and robotics. A special focus is given to the Code 39 symbology, examining how its unique technical characteristics---such as variable length, self-checking properties, and low data density---influence its ongoing use in these diverse sectors. |
|
1. Introduction: The Gift of Sight for Machines |
The modern industrial landscape is a testament to human ingenuity, a complex ecosystem of factories, warehouses, and logistics networks that operate with incredible speed and precision. Yet, for all their automation, these systems have long suffered from a critical sensory gap: they have been blind. While machines could perform repetitive tasks with tireless mechanical accuracy, they lacked the ability to perceive and interpret their environment. This is where machine vision enters the picture. |
Machine vision is the technological enabler that grants automated systems the ability to see. It is the integration of cameras, processors, and sophisticated algorithms into a single system designed to capture and interpret images for the purpose of automated inspection, guidance, and control . It transforms a camera from a simple image-capturing device into an intelligent sensor that can make decisions based on what it sees. In the world of industrial automation, machine vision is the eye of the machine. |
This technology is not to be confused with the broader, more theoretical field of computer vision. While both involve the processing of visual data, computer vision is an academic discipline focused on computational theories and algorithms that enable machines to understand images, often with a focus on mimicking human visual cognition . Machine vision, on the other hand, is the practical application of these principles. It is an engineering discipline focused on solving specific, well-defined industrial tasks. It is robust, high-speed, and designed to operate in the harsh and demanding environments of a factory floor, reading barcodes on a package flying down a conveyor belt or guiding a robotic arm to pick up a part . |
The evolution of barcode scanning is a perfect illustration of the power of machine vision. Traditional barcode scanners, often laser-based, are limited to reading codes by scanning a laser beam across the code. They require the code to be oriented in a specific direction and are easily confused by print distortions or poor contrast. Machine vision, however, uses cameras to take a picture of the entire scene. An integrated processor then analyzes the digital image using complex algorithms to locate and decode any barcode present, regardless of its orientation (omnidirectional reading), even if it is slightly distorted, torn, or placed on a curved surface . This robustness and flexibility have made machine vision the backbone of modern tracking, tracing, and automation systems, with one prominent example being its use in logistics: every package that makes its way through today's automated logistics warehouses does so with the help of barcodes that are read by machine vision . |
|
2. The Core Components of a Machine Vision System |
To understand how machine vision works, it is helpful to break down its architecture. A typical industrial machine vision system consists of several key components that work in harmony to capture and interpret the visual world. |
The Camera and Sensor |
The camera is the fundamental 'eye' of the system. Unlike standard consumer cameras, industrial cameras are designed for reliability and precision. They are often equipped with image sensors that have high resolution to capture fine details, such as small barcodes or microscopic defects. These sensors, often CMOS or CCD, convert light that falls onto them into an electronic signal. The choice of sensor and camera features---like global shutters to avoid motion blur---is critical for capturing clear and usable images of fast-moving objects . |
Lighting |
Lighting is arguably the most critical component of any machine vision system. The old adage in computer vision, 'garbage in, garbage out,' is particularly true here. The quality of the image captured is directly dependent on the quality of the lighting. A well-designed lighting scheme creates the necessary contrast to highlight the features of interest while suppressing irrelevant details or reflections. For example, red light is often used to illuminate barcodes printed in red, while blue light can be used to improve contrast on stainless steel surfaces . Various lighting techniques, such as backlighting to create silhouettes for dimensioning, ring lights to provide even illumination, or polarizers to eliminate glare from shiny packaging, are all essential tools in the machine vision engineer's arsenal . |
The Processor and Software |
The processor and software constitute the 'brain' of the machine vision system. While the camera and lens capture the image, the processor---often a dedicated industrial computer or an embedded system---performs the heavy lifting. This is where the 'vision' part of machine vision happens. The software is the core of the system, running powerful algorithms to extract information from the image. |
There are two primary approaches to machine vision software : |
Rule-based (or Traditional) Machine Vision: This approach relies on a set of user-programmed step-by-step instructions. The user defines specific criteria, such as the expected location of a barcode, its physical dimensions, and its contrast. The software then uses this 'recipe' to locate and decode the code. This method is excellent for consistent, high-speed tasks where the application is highly controlled. |
AI-Powered Machine Vision: This approach employs artificial intelligence, often in the form of neural networks. Instead of a programmer defining rules, the software is 'trained' on a large database of labeled images. It learns to identify patterns and features, making it exceptionally robust to variability. AI-powered systems are ideal for handling variations in lighting, orientation, distortion, and even damaged codes that would challenge a rule-based system . They are also used for more complex tasks beyond reading codes, such as detecting subtle product defects. |
Many modern systems combine both rule-based and AI-powered methods to provide the most efficient and robust solution . |
|
3. Machine Vision in Action: Across the Industrial Spectrum |
The power of machine vision lies in its versatility. Its ability to provide high-speed, accurate, and reliable visual information has led to its adoption across virtually every sector of industry. The following are some of the most prominent areas where machine vision is making a tangible difference. |
Logistics and Warehousing: The Backbone of Modern Commerce |
The logistics industry has become the lifeblood of the global economy. The rapid movement of millions of packages each day requires levels of speed and accuracy that are impossible for human workers to sustain. Machine vision is the core technology that makes this possible. |
In a modern automated warehouse, packages of all shapes and sizes zip down conveyor belts at breakneck speeds. Machine vision cameras, strategically placed along the line, capture images of the packages to identify them. They read the barcodes---whether 1D, 2D, or even direct part marks---on every surface of the package without needing it to be oriented in a specific way . This is a dramatic improvement over older systems where a worker had to manually orient a package to face a scanner. |
The application goes far beyond simple reading. The cameras are used for dimensional scanning, measuring the size of a package to ensure it fits within a shipping container and to calculate shipping costs. They are integrated with sortation systems; by reading a barcode, the system instantly knows a package's destination and directs it to the correct chute for a particular truck or route . This creates an unbroken chain of traceability from the moment a package enters the system until it is delivered to its final destination. |
Furthermore, the latest systems leverage AI-powered detection to manage exceptions and complex tasks. Companies like Purolator have implemented Cognex's SLX logistics portfolio, which combines barcode reading with AI-powered item detection . This allows the system to not only read the code but also verify that the package is there, classify it, and ensure it is not damaged, all in a single automated step. |
Manufacturing: Quality Control and Beyond |
In manufacturing, machine vision is synonymous with quality control. The technology is used at every stage of production to ensure products meet strict quality standards, thereby reducing waste and preventing defective products from reaching customers . |
Defect Detection and Inspection: Machine vision can inspect components at microscopic scales, far exceeding human capabilities. In the electronics industry, for example, it can detect soldering defects on a circuit board or find tiny scratches on a semiconductor . In the automotive sector, it can check for flaws in a stamped metal part. This automated inspection is not only more accurate and faster than human inspection but also objective and consistent . |
Process Control and Guidance: Machine vision guides robots to perform complex tasks. In bin-picking, a 3D camera captures a pile of random parts, and the machine vision system determines the orientation and location of a part that a robot can grip . The system can also read the barcode or data matrix code on the part simultaneously in the same capture cycle, ensuring that the robot picks the correct part from a mixed bin and linking the physical part to its digital record for traceability . |
Full Traceability: Traceability is a critical need in industries like automotive, aerospace, and medical devices. If a part fails, it is essential to trace it back to its raw material batch and the specific machine and time it was produced. Machine vision is the lynchpin of this process, reading direct part marks and barcodes to log every step a part takes in its production lifecycle, creating a rich data set for analysis . |
Healthcare and Pharmaceuticals: Patient Safety and Efficiency |
The healthcare and pharmaceutical industries operate under some of the most stringent regulations in the world, as errors can have life-or-death consequences. Machine vision plays a vital role in ensuring patient safety and operational efficiency . |
Asset and Sample Tracking: In hospitals and labs, machine vision is used to label and track everything from patient wristbands to vials of blood or test strips. A patent application for a point-of-care diagnostic system details how a Code 39 barcode on a test strip cassette can encode information like the lot number, expiration date, and the specific analyte being tested for . Reading this code ensures that the correct test is run for the patient, the test is performed with the correct reagents, and expired materials are not used . |
Pharmaceutical Packaging: On a pharmaceutical packaging line, machine vision cameras inspect blister packs for pills. They verify that each pocket contains a pill, that the pill is not broken, and that the coloring is correct. They also read barcodes on the packaging to ensure the right labels are on the right boxes, a critical step in preventing medication errors. |
Inventory Management: As discussed in a recent application note, edge AI-powered systems are being used for inventory checks in pharmacies and hospitals . These systems use cameras to perform both optical character recognition (OCR) to read text and barcode recognition simultaneously to validate product information and expiration dates in real-time, eliminating manual visual checks and reducing the risk of human error . |
Food and Beverage: Safety and Freshness |
In the food and beverage industry, quality and safety are paramount. Machine vision ensures that the products we consume are safe, correctly labeled, and fresh . |
Traceability and Recalls: Track-and-trace capabilities are crucial for managing product recalls. Machine vision reads barcodes on every package to create a digital record of its journey from production to sale . If a contamination issue is discovered, the manufacturer can quickly pinpoint the specific batch and time of production, allowing them to issue a targeted recall rather than pulling all products from the market. |
Package Integrity and Inspection: Vision systems inspect food packaging at various stages, ensuring that labels are correctly applied, 'Best Before' dates are clearly printed, and seals and tamper bands are intact . In bottling applications, they check fill levels to ensure consumers are getting the correct amount and that caps are properly sealed to maintain freshness . |
Quality Assurance: The technology can even inspect the food itself. Advanced systems can use spectral imaging to inspect fruit for bruising or determine ripeness, ensuring that only the highest quality products reach the store shelves . |
|
4. Code 39: A Foundational Symbology in the Age of Machine Vision |
In the landscape of barcodes, Code 39 holds a special place. While newer and more compact symbologies like Code 128 and two-dimensional codes like Data Matrix have been developed, Code 39 remains a widely used and highly relevant standard, particularly in certain industrial sectors . Its prevalence is directly tied to its historical significance, its simple and robust encoding, and its technical characteristics, which bring both advantages and limitations in the context of machine vision. |
A Brief History and Definition of Code 39 |
Code 39, also known as Code 3 of 9, was developed in 1974 by Intermec Corporation . It was a breakthrough in its time, as it was the first barcode symbology that could encode both numbers and letters, not just numeric digits . This capability to encode alphanumeric data made it instantly useful for a much wider range of industrial applications. It is a discrete, variable-length symbology, meaning that it can encode any number of characters and each character is encoded independently, without relying on its neighbors for interpretation . This characteristic simplifies the encoding and decoding process. |
Technical Characteristics of Code 39 |
To understand why Code 39 is still used, we must examine its core technical features. |
The Encoding Scheme: 5 Bars and 4 Spaces |
The name 'Code 39' comes from its foundational encoding scheme. Each character in the Code 39 character set is represented by a pattern of nine elements: five bars (the dark lines) and four spaces (the light gaps between them) . Of these nine elements, exactly three are wide, and the other six are narrow . This '3 of 9' pattern is what gives the code its name. The ratio of the width of a wide bar to a narrow bar is typically between 2.5:1 and 3.0:1 . This simple and consistent pattern makes the code relatively easy to print and decode. |
Character Set: The Alphanumeric Enabler |
The standard Code 39 character set can encode a total of 43 characters: the ten digits (0-9), the 26 uppercase letters (A-Z), and seven special characters: space, period (.), minus (-), plus (+), dollar sign ($), slash (/), and percent (%) . The asterisk (*) is used exclusively as a start and stop character to indicate the beginning and end of the barcode and is never intended to be a data character . |
A further extension, known as Code 39 Extended or Full ASCII Code 39, was developed to encode the entire 128-character ASCII set, including lowercase letters . It accomplishes this by using two-character combinations from the standard Code 39 character set to represent a single ASCII character. For example, lowercase 'a' is represented as '+A'. While this provides full ASCII support, it effectively doubles the length of the barcode for those characters, reducing the already low data density even further . |
Self-Checking Property |
One of the most significant technical advantages of Code 39 is its self-checking property . Because each character is composed of a pattern of exactly three wide and six narrow elements, a single print defect---like an ink blob that widens a narrow bar or a scratch that narrows a wide one---is highly unlikely to transform a valid character into another valid character. The pattern is simply too different. If a character is corrupted, the decoder will recognize that it doesn't adhere to the 3-of-9 rule and reject it as an invalid character. This gives the symbology a degree of inherent error resistance without the need for a mandatory checksum character . |
Low Data Density |
This simplicity and robustness comes with a significant trade-off: very low data density . Because each character requires nine elements of varying width, the barcode must be quite long to encode even a modest amount of data. A 10-character Code 39 code can be roughly 40% wider than an equivalent Code 128 code . This means that Code 39 is not suitable for applications where space is at a premium, such as small electronic components, where a compact 2D code like Data Matrix would be necessary. However, it is perfectly adequate for larger labels, such as those on shipping boxes or automotive parts. |
Optional Checksum |
While Code 39 is self-checking, it does not have a mandatory checksum character. It can be, and often is, used without one. However, for applications requiring a higher level of security, an optional checksum digit, calculated using a Modulo 43 algorithm, can be appended to the data before the stop character . Many industry standards, like the U.S. Department of Defense MIL-STD-130, mandate the use of this Modulo 43 checksum for all Code 39 applications . |
How Code 39's Technical Features Shape Its Application Across Industries |
The unique combination of Code 39's features---its variable length, low data density, self-checking property, and optional checksum---has dictated its use across different sectors. In the age of machine vision, these characteristics don't make the code obsolete, but rather define its niche. |
Automotive Industry |
The automotive industry is one of the largest and most persistent users of Code 39. The Automotive Industry Action Group (AIAG) established its standard B-1 for parts labeling, which was originally based on Code 39 . This standard is deeply embedded in the automotive supply chain. A major automotive manufacturer or supplier cannot simply decide to switch to Code 128 without requiring a massive and costly change across all their suppliers' labeling systems. In the world of machine vision, this means that a high-quality automotive application must be designed to read Code 39 barcodes, even in challenging conditions. Code 39's self-checking property makes it relatively resistant to the dirt, grease, and wear that are common on metal auto parts in a stamping or welding environment, and its variable length allows it to encode a VIN or a part number of varying sizes . |
Defense and Government: The LOGMARS System |
The U.S. Department of Defense (DoD) is a key reason Code 39 became a widespread standard. They developed LOGMARS (Logistics Applications of Automated Marking and Reading Symbols), which mandated the use of Code 39 for all government property marking . Like the automotive industry, the DoD's massive logistics network is standardized on Code 39. Suppliers to the U.S. military must, by contract, use Code 39 barcodes. Machine vision systems used in these defense supply chains must be incredibly robust. They must be able to read these codes from a distance, often in poor lighting, and under strict security and compliance regulations. |
Healthcare |
The healthcare industry, particularly through the Health Industry Bar Code (HIBC) standard, also makes extensive use of Code 39 . In hospitals, Code 39 barcodes can be found on everything from intravenous (IV) bags to patient wristbands to medical devices. Here, the variable length is useful, as different items require different amounts of information. The self-checking property and optional checksum provide a necessary layer of safety against misidentification that could lead to a medical error. As seen in a patent for a point-of-care diagnostic device, Code 39 was chosen for its suitability in encoding essential test and patient data on a cassette . The machine vision systems in a hospital setting must be extremely reliable and require high read rates to ensure the right medication is given to the right patient. |
Internal Asset Tracking |
The simplicity and ease of printing Code 39 barcodes has made them a popular choice for internal asset tracking in libraries, equipment management, and document routing . Since the code does not require a complex checksum to be generated for printing, it's simple for organizations to create in-house asset labels. In the era of machine vision, these internal systems are also being upgraded with cameras. An employee can use a machine vision-enabled handheld computer to scan a Code 39 barcode on a microscope or a laptop in an office, instantly pulling up the asset's maintenance history or software inventory. |
Logistics |
While Code 128 is more common on consumer packages due to its higher data density and ability to encode more information in a smaller space, Code 39 is still used extensively in logistics for larger shipping labels . The logistic industry's machine vision systems, like the previously mentioned SLX devices, are designed to 'read any symbology,' and that includes Code 39 . The ability to decode a Code 39 barcode from a meter away, regardless of the orientation of the box on a high-speed conveyor, is a testament to the flexibility of modern machine vision. |
|
5. The Future of Machine Vision: AI, Edge, and 3D |
Machine vision is not a static technology; it is constantly evolving. The integration of cutting-edge technologies is expanding its capabilities, making it more intelligent, flexible, and ubiquitous. |
AI and Deep Learning |
As mentioned in the core components section, AI is revolutionizing machine vision. Rule-based systems are excellent for predictable tasks but struggle with variability. When a barcode is damaged, partially occluded, printed on a reflective surface, or on a label with a completely different layout, a rule-based system might fail. AI-powered systems, however, are trained to handle this 'noise.' They learn to recognize a barcode even when it's not perfect. For example, a Vision AI Label Reader used in electronics logistics can handle different label layouts, languages, and even damaged codes without needing reprogramming . This flexibility makes AI the future of automated inspection and code reading. |
Edge AI |
The concept of 'edge computing' is a natural partner for AI in machine vision. Edge AI means that the AI processing happens on the device itself (the 'edge'), rather than sending the image data to a remote cloud server . This has several critical advantages: |
Speed: Processing is done in real-time, with low latency, allowing for immediate decisions. |
Reliability: It is not affected by network outages or slow connections. |
Security: Sensitive visual data never leaves the local device, which is crucial for patient data in healthcare or proprietary designs in manufacturing . |
Systems like the one described by Macnica demonstrate this principle, where a single system-on-chip (SoC) performs both OCR and barcode recognition on an edge device for real-time inventory management . |
3D Vision |
For decades, machine vision was largely a two-dimensional technology. However, the physical world is 3D, and robots need to interact with it in three dimensions. The integration of 3D cameras and software is unlocking new capabilities . In bin-picking, a 3D camera captures the depth information needed for a robot to pick an object. The Zivid technology example shows how a 3D camera can now also perform barcode reading at the same time . This is a major advancement in robotic workflows: |
- In order picking, the robot identifies the item's 3D position and reads its barcode in one step, eliminating the need for a separate scan, saving 5-10 seconds per pick and reducing the complexity of the work cell . |
- In palletizing, a robot can read a barcode on a box while also computing its precise pose for stacking. This allows the robot to intelligently sort boxes for different orders or destinations based on the barcode data . |
By merging 2D image analysis with 3D spatial awareness, machine vision is giving robots the complete sensory picture they need to operate with true autonomy. |
Connected and Integrated Systems |
Finally, the future of machine vision lies in connectivity. Data generated by machine vision systems is no longer just used for a single, isolated task like rejecting a part. It is fed into manufacturing execution systems (MES) and enterprise resource planning (ERP) systems, creating a wealth of data for analysis . This 'big data' can be analyzed to optimize the entire production process, identify bottlenecks, and even predict machine failure before it happens. As the technology becomes more connected, it becomes less of an inspection tool and more of a central nervous system for the entire industrial operation. |
|
6. Conclusion: A Detailed Summary of Machine Vision and Code 39 |
Machine vision has fundamentally transformed the way modern industry operates. It is the technology of seeing and understanding, a crucial bridge between the physical world of manufacturing and the digital world of data. By combining cameras, optics, lighting, and advanced software---increasingly powered by AI---machine vision provides automated systems with a level of perception that rivals, and in many cases surpasses, human capabilities. |
Its applications are vast and vital. In logistics, it is the silent workhorse that powers the sorting and tracking of billions of packages, ensuring that the global supply chain functions smoothly . In manufacturing, it ensures product quality by identifying microscopic defects, guides robots to perform precise assembly, and provides the traceability needed for compliance in high-stakes industries like aerospace and medical devices . In healthcare, it safeguards patient safety by ensuring the right medication is given to the right patient and that medical devices are free from defects . In the food and beverage industry, it protects the public by verifying labels, detecting contaminants, and providing crucial traceability for recalls . |
Within this ecosystem of barcode reading, Code 39 stands as a foundational and enduring technology. Born in 1974, it was the first symbology to bring alphanumeric data to barcodes, and its impact is still felt today . Its technical characteristics define its role. |
The '3 of 9' encoding scheme is simple, making the code easy to print and decode . |
Its self-checking property provides inherent resistance to minor print defects, a major advantage in industrial environments where labels can become dirty or damaged . |
Its variable length offers flexibility, allowing for the encoding of data fields of different sizes, from a few digits to a long alphanumeric VIN . |
Its optional Modulo 43 checksum provides an additional layer of security for applications where data integrity is critical, such as defense and healthcare . |
Its primary limitation is its very low data density, which means it requires a large amount of space to encode a modest amount of data, making it unsuitable for small items . |
These characteristics have made Code 39 a standard in specific industries. The automotive industry remains anchored to it via the AIAG standard . The U.S. Department of Defense mandates it for all government property marking through the LOGMARS system . The healthcare industry uses it widely through HIBC standards . Despite the rise of more modern, space-efficient symbologies like Code 128 and 2D codes, Code 39 persists because of the immense cost and logistical difficulty of changing deeply embedded industry standards. Its reliability and simplicity ensure that it will remain a valid and frequently encountered code in machine vision applications for the foreseeable future. |
|
The evolution of machine vision continues at a rapid pace. The integration of AI and deep learning is making systems more robust and flexible, enabling them to handle the variability of the real world with unprecedented ease . The rise of Edge AI is bringing this intelligence directly to the point of action, enabling real-time decision-making with unmatched speed and security . The addition of 3D vision is giving machines not just sight, but depth perception, empowering robots to perform complex tasks that were once impossible . |
Machine vision has moved from a niche technology to an essential pillar of modern industrial civilization. Its ability to provide speed, accuracy, and objectivity is not just improving efficiency; it is enabling entirely new paradigms of automation. As the technology continues to become more powerful and intelligent, it will undoubtedly become an even more pervasive and invisible force, seamlessly guiding the machines that build and deliver the world around us. |
cs@easiersoft.com