Chapter 68: The 'Unreadable' Becomes Readable |
In Brief |
This chapter explores a fundamental shift in machine vision: the ability to read barcodes that have been physically unreadable due to surface conditions. Traditional systems rely on print contrast---the difference between light and dark areas. Photometric stereo and 3D reconstruction bypass this limitation by reading the geometry of the code itself, not merely its color or ink. We will examine how this works, its implications across multiple industries, and how the enduring Code 39 symbology intersects with these emerging capabilities. |
|
1. The End of the Print Contrast Era |
For the past fifty years, the reliability of a barcode scan has been largely dependent on a single factor: print contrast. A standard barcode is a simple optical pattern. A scanner shines a light on the label; the black bars absorb the light, and the white spaces reflect it back. A sensor detects this variation in reflected light and translates it into the ones and zeros of data . This system works exceptionally well on cardboard boxes, glossy labels, and paper documents. It fails, however, the moment the 'white' becomes gray, the 'black' fades, or the surface becomes reflective. |
In heavy industry, labels burn, fade, or get covered in grease. In manufacturing, barcodes are often not printed at all. They are etched, embossed, or dot-peened directly onto metal. In these instances, there is very little contrast. A dot-peened code on a steel piston might look like a series of tiny indentations that are the same color as the surrounding metal. A traditional optical scanner sees little to no difference between the 'bar' and the 'space' because the color is identical. |
Future machine vision systems are moving beyond the limitations of 2D contrast. They are entering the realm of the third dimension. By using techniques like photometric stereo, these systems can detect microscopic bumps, pits, and grooves in a surface. They read the shape of the code rather than the color of the code. The 'unreadable' becomes readable because the system no longer cares about the ink; it cares about the topography. |
|
2. The Technology of Topography: Photometric Stereo and 3D Reconstruction |
To understand how we read what was once invisible, we must first understand the core technology enabling this shift: Photometric Stereo. |
What is Photometric Stereo |
Photometric Stereo is a technique for deriving the surface orientation of an object by observing it under different lighting conditions . While a standard 2D camera takes one picture with one light source, a photometric stereo system takes multiple images of the same object from the same camera position, but with lights positioned in different directions. |
Imagine a pebble on a flat surface. If you shine a light from the left, the left side is bright, and the right side is cast in shadow. If you shine a light from the right, the opposite happens. By analyzing how the brightness of each pixel changes as the light moves, the computer can calculate the slope, or orientation, of that tiny point on the surface . It builds a 'normal map'---a digital representation of which way every point on the surface is facing . |
These normal maps, along with calculated data like mean and Gaussian curvature, are powerful. They allow the system to 'flatten' the image in the software, reconstructing a 3D representation that reveals the embossed or etched features that the human eye might miss . |
|
Why This Matters for Barcodes |
For a barcode to be read by these new systems, it no longer needs to be black and white. If a code is etched into metal, the indentation creates a curvature. The walls of the etch reflect light differently than the flat surface. Photometric stereo detects these changes in orientation and creates a synthetic grayscale image where the 'highs' and 'lows' of the surface appear as black and white contrasts . |
This is a massive leap forward for industries where durability is paramount. A code that is painted onto a rough surface will eventually be scratched off; a code that is embossed into the surface will remain for the lifetime of the part. With photometric stereo, a machine vision camera can look at a rough, cast-iron engine block and read the dot-peen code with high reliability, independent of whether the code has rusted or been painted over . |
The Challenge of Shiny Surfaces |
It is important to note that photometric stereo is not a magic cure-all. It works best on matte surfaces because they reflect light diffusely---meaning they bounce light back in all directions . Highly reflective, specular surfaces (like polished metal) can cause issues because they act like mirrors and reflect light in a single direction, creating bright spots that confuse the algorithm . However, modern implementations are beginning to solve these challenges through better calibration, multi-spectral lighting, and algorithms designed to detect and remove highlights . |
|
3. Code 39: The Enduring Workhorse |
Before we look to the future, it is necessary to understand the present. One of the most common symbologies in the world, and the one most likely to be found in these harsh industrial environments, is Code 39. |
What is Code 39 |
Code 39, also known as 'Code 3 of 9,' was invented in 1974 by Intermec. It was revolutionary for its time because it was the first barcode symbology capable of encoding alphanumeric characters (letters and numbers) in addition to just digits . It is a discrete, variable-length symbology. Each character is encoded by a pattern of five bars and four spaces, totaling nine elements. Of these nine elements, three are wide and six are narrow---hence the name '3 of 9' . |
Key Characteristics of Code 39 |
1. Character Set: Code 39 supports 43 characters: uppercase letters (A-Z), numbers (0-9), and several special characters (space, ., -, /, %, $, +) . |
2. Self-Checking: Code 39 has an inherent property called 'self-checking.' Because each character has a specific pattern of wide and narrow elements, a single substitution error (like a wide bar being misread as narrow) will usually result in an invalid pattern, preventing a misread . |
3. Check Digit: It does not require a check digit. While an optional 'Modulo 43' checksum can be added, it is often omitted, which reduces data integrity in high-stakes scenarios . |
4. Low Data Density: This is the primary limitation of Code 39. Because it uses nine elements per character and wide elements take up more space, the barcode is very long. It is one of the lowest-density barcodes. Encoding the word 'CODE' might take several inches of label space . |
5. Wide Support: Code 39 is supported by practically every barcode reader ever manufactured. It is the 'lingua franca' of industrial scanning . |
How Code 39's Characteristics Influence Its Use |
The 'Logmars' Military Standard: Code 39 was adopted by the US Department of Defense for the LOGMARS (Logistics Applications of Automated Marking and Reading Symbols) system. This forced suppliers to use Code 39 on military equipment, cementing its place in defense and aerospace . |
Automotive and Manufacturing: It is heavily used in the automotive industry for tracking parts (like VIN numbers and assembly components) because of its alphanumeric flexibility and the robust nature of its wide/narrow pattern . |
Healthcare: It is used for specimen labeling and patient identification, often via the HIBCC (Health Industry Bar Code) standard . |
The Size Limitation: Because Code 39 is low-density, it is rarely used in retail (where space on a small item is limited). Instead, it is used in logistics and warehousing, where labels can be large, and the scanning distance is variable. The size helps manual operators scan it easily from a distance . |
|
4. Industry Applications: Where the Unreadable Becomes Readable |
The convergence of 3D imaging techniques with traditional symbologies like Code 39 is transforming heavy industries. Here is a look at how these technologies work in practice. |
1. Aerospace and Defense: The 'Return on Investment' of Reliability |
In aerospace, failure is not an option. Parts are made of expensive alloys like titanium and Inconel, and they have to last for decades. Barcodes on these parts are often dot-peened or laser-etched directly onto the surface. These codes provide traceability back to the casting, the batch of material, and the quality control tests performed . |
The Problem: Traditional cameras struggle to read dot-peened codes because the contrast is very low. Depending on the angle of the light, the dots might appear as shadows or not appear at all. |
The Application: A European aerospace manufacturer recently struggled to read 2D dot-peened barcodes with standard technology. They implemented a system using Photometric Stereo techniques deployed on a powerful smart camera. Instead of one image, the system took several images under different lighting angles. The software composed a 'pseudo-3D' image that highlighted the height differences of the dots. The code, previously unreadable, became perfectly clear, allowing for seamless traceability . |
2. Steel and Heavy Manufacturing: Reading the Red Glow |
Steel production is one of the harshest environments on the planet. Slabs of steel move through rolling mills at high speeds and extreme temperatures. They are covered in a layer of mill scale (iron oxide) that is rough and dark. |
The Problem: A slab might have a painted barcode or a stamped code. However, the heat causes the paint to flake off, and the scale builds up over the stamped code. A standard 2D system cannot see the difference between the code and the scale. |
The Application: Researchers have developed multiview multispectral photometric stereo instruments to detect surface defects on hot continuous casting slabs . While the primary goal is defect detection (scars, cracks, indentations), the same technology provides the 'topography' required to read stamped codes. Because the system detects depth variations (high-frequency information of surface depth variations), it can read an identification number that is physically part of the steel, even while the steel is glowing red at temperatures that would destroy a paper label. This system is fast enough to work online, processing images in less than 3.6 seconds . |
3. Logistics and Warehousing: The Curved and the Crumpled |
Logistics is the traditional heartland of barcode scanning, but even here, 3D techniques are making inroads. |
The Problem: Packages are often crushed, wrapped in uneven plastic shrink-wrap, or placed on curved surfaces like pipes or bottles. When a barcode is bent around a cylinder, the geometry of the bar code is distorted. The wide spaces look thin, and the thin spaces look wide. A standard 2D reader fails to decode this because it is looking for straight lines . |
The Application: While photometric stereo is excellent for etched metal, simpler 3D reconstruction algorithms are being used in mobile devices and AI-driven cameras to handle curved surfaces. By detecting the curvature of the label itself, software can mathematically 'unwrap' the barcode virtually, flattening it so that the scanner sees it as a flat plane. This is particularly useful for scanning QR codes on food trays or tubular surfaces . AI-powered scanning tools are increasingly capable of reading these distorted barcodes, which is vital for reverse logistics (handling returns) where the packaging is rarely pristine . |
4. Automotive and Foundries: Castings and Embossing |
In an automotive foundry, parts like cylinder heads, engine blocks, and transmission housings are cast from molten metal. |
The Problem: These parts have rough, 'as-cast' surfaces. If an identification is embossed (raised) onto the surface during the casting process, the letters or bars might be the exact same color as the rough background. A traditional light-based scanner cannot differentiate between the raised letter and the rough sand texture. |
The Application: Systems using photometric stereo extract the 'mean curvature' of the surface. The raised letters have a different curvature than the surrounding rough metal. By isolating the Gaussian or mean curvatures, the system reads the code easily . This also applies to Braille and embossed text on pharmaceutical packaging, where the contrast is minimal . These systems are often combined with robot-guided cameras to navigate complex geometries on large castings . |
5. Medical and Pharmaceutical: Tiny Codes on Tiny Parts |
In medical devices, parts are getting smaller and smaller. Stents, surgical instruments, and implantable devices require traceability. |
The Problem: There isn't enough room on a surgical screw for a large Code 39 label. Usually, a 2D Data Matrix code is used, which is tiny. These codes are often laser-etched. |
The Application: Photometric stereo allows vision systems to read these tiny etched codes regardless of the surface finish. Even if the part is polished (specular), the depth of the etching provides enough curvature for the system to create a readable image . This ensures that a surgical screw can be traced back to its manufacturer and batch, even after it has been sterilized (which can fade ink). |
|
5. The Role of AI and Deep Learning |
Traditional barcode reading relies on fixed algorithms that look for specific patterns of bars and spaces. These algorithms often fail when the code is skewed, damaged, or poorly illuminated. |
This is where Artificial Intelligence (AI) and Deep Learning are making a significant impact, particularly in synergy with 3D imaging. |
Pre-Trained Networks: Modern tools like Cognex's 'Blue Read' or Zebra's AI-based OCR are pre-trained on millions of images. They don't just look for black and white; they understand the structure of a character or a barcode. When an etched character is partially obscured, the AI can infer what the full character should be based on its training . |
Handling Complex Surfaces: AI helps define the Region of Interest (ROI) on a complex, embossed surface. By using curvature data from photometric stereo, the AI can isolate the specific location of the code and 'ignore' the surrounding noise . |
Robustness: AI-powered scanners can read barcodes that are up to 25% more damaged than those readable by traditional scanners. They handle blur, skew, and uneven lighting better than algorithmic approaches . |
Speed: Modern AI engines can process up to 500 barcodes per minute, allowing for fast-moving production lines to maintain traceability even when using complex imaging techniques . |
|
6. A Summary: The View from the Future |
In this chapter, we have seen a paradigm shift. The fundamental constraint of 'Print Contrast' is being eliminated by the application of 'Depth Geometry.' The barcode is no longer a visual marking placed on top of a surface; it is becoming a physical property of the surface itself. |
Photometric Stereo allows systems to see the geometry of a code by analyzing how it reflects light from multiple angles. |
3D Reconstruction allows systems to 'flatten' codes that are curved or deformed. |
AI and Deep Learning enhance this data, allowing the system to infer the correct data even when the visual image is imperfect. |
|
The Future of Code 39 |
Despite the innovations in decoding, Code 39 will not disappear. It is too deeply entrenched in the industrial ecosystem. However, its relationship with reading technology is changing: |
1. The 'Safety Net': In aerospace, military, and automotive sectors, where Code 39 is the standard, the use of photometric stereo ensures that the labels that have faded or eroded over the years can finally be read. This saves time and reduces the human error of manual data entry. |
2. The 'Legacy' Factor: For the thousands of factories that rely on Code 39, upgrading to a 3D vision system does not require changing their labels. They can keep their existing stampers and embossing tools. The new camera systems simply have the intelligence to 'see' them better. |
3. Increasing Reliability: While Code 39 is self-checking, it does not have a mandatory check digit. The addition of AI-powered reading logic, combined with 3D detection, effectively creates a 'digital check digit.' The system confirms the physical geometry before it even tries to decode the pattern, reducing the theoretical misread rate to near-zero. |
4. The Rise of 2D: However, it is clear that the future of high-density data lies with 2D codes like QR Codes and Data Matrix. They store more data in less space and have built-in error correction. In medical and electronic manufacturing, 2D codes are already dominant. The technology described in this chapter (photometric stereo) is often used to read these 2D codes on the smallest, most difficult surfaces . |
In conclusion, the future of machine vision is not about brighter lights or higher resolution cameras. It is about smarter light, intelligent geometry analysis, and a willingness to stop looking at the surface and start looking into it. Whether it is a Code 39 barcode stamped on an aircraft wing or a QR code embossed on a hypodermic needle, the 'unreadable' has finally become readable. The limiting factor is no longer the condition of the surface; it is merely the presence of physical structure. |
This marks the dawn of a truly reliable, truly traceable industry, where every part, regardless of how long it has been in service or how harsh its environment has been, can be identified and verified instantly and accurately. The technology transforms the barcode from a visual art to a physical attribute, ensuring that data is as permanent as the material it resides on. |
cs@easiersoft.com