• Font Size:
  • A
  • A
  • A

Tech Papers

Five Signs that a Photometry-Based Imaging System is the Right Choice for Your Inspection Application

Radiant Vision Systems

Introduction

Human visual perception and judgment is the standard when inspecting parts for low-contrast, subtle, and unpredictable defects. In industries like consumer electronics, aerospace, and automotive manufacturing, human inspectors may be stationed on the line to perform complex assembly verification or surface inspection. These tasks can range from inspecting the internal components of a laptop assembly to checking for cosmetic defects on smart device housings to evaluating the finish on vehicle surfaces. In these industries where perfection is the expectation, the smallest defect can spell disaster for a manufacturer’s reputation. These defects are also the hardest for automated systems to detect.

Man holding up smartphone for inspection

Humans have remained on the production line despite operational inefficiencies thanks to their unique ability to detect subtle, unpredictable, and contextually-dependent errors beyond the capabilities of an automated system.

Automated visual inspection by machine vision is superior to human inspection in terms of objectivity, efficiency, and repeatability. However, humans have remained on the production line – despite operational inefficiencies – thanks to their unique ability to detect and assess errors that are too subtle, too unpredictable, or require context too specific to be digitally programmed on an automated system. Therefore, where a flawless finish and error-free operation is essential, many makers of high-value products continue to rely on manual inspection methods.

There may be a suitable medium for manufacturers who require the visual acuity of human inspectors but desire the benefits of automation. Imaging photometers, (CCD-based cameras designed to measure light intensity and luminance as it is perceived by the human eye), can be combined with production sequencing software and adapted to a wide range of complex inspection tasks. These cameras feature high resolution and broad dynamic range that enables precision inspection in applications where standard machine vision solutions often fall short. The five indicators outlined in this paper are intended to help readers understand the capabilities of a photometry-based imaging solution and its suitability for advanced inspection applications.

The Limitations of Standard Machine Vision

A standard machine vision camera, (which can be defined as a two-dimensional radiometric imager comprised of a CCD or CMOS sensor, commonly with a resolution of 12MP or less), is ideally suited to applications with visually obvious and predictable defects, with few anomalies or defects that fall out of the scope of programmed points of interest. These systems will generally sacrifice precision for speed, offering comparatively low-resolution sensors that capture just enough image detail to detect a named error. Programmed to analyze image features within defined points of interest, standard machine vision system software tools are generally applied in static locations to look for defects like a screw missing from a drill hole, or an incorrectly-positioned label on a product. If shapes, sizes, colors, or locations of part features change, the machine vision system may be unable to detect the defect amid the range of contrast variations in the image.

For this reason, these systems are typically applied in highly repetitive applications, where image variations can be easily augmented with strategic lighting, and little judgment or contextual information is needed for analysis. The relative ease of visual inspection in these applications makes machine vision a good alternative to human inspection, providing a repeatable, automated solution for quality control. For more complex applications, however, a standard machine vision system may not offer the imaging and software capability required for 100% defect detection.

Radiant Vision Systems Camera

Imaging photometers feature high-resolution sensors and advanced optical components to capture images in greater detail than a typical machine vision camera.

Photometry-Based Imaging Systems

An alternative solution to standard machine vision that offers superior imaging and software functionality is a photometry-based imaging system. Commonly used to test and measure light-emitting products such as displays and LED lighting, imaging photometers are also beneficial in complex inspection applications of non-illuminated devices. The hallmark of imaging photometers is that they are designed to weigh the intensity and power of light with respect to the sensitivity of the human eye. Since visual inspection relies on interpreting reflections or emissions of light to capture and process images, a system that more closely simulates human perception of light is more capable of performing human-like error detection.

Imaging photometers also feature much higher-resolution sensors and advanced optical components to capture images in greater detail than a typical machine vision camera. This allows more subtle defects to be detected and, therefore, more functions to be applied to meaningful variations within an image. Combined with advanced photometry-based software tools, imaging photometers interpret these details to capture anomalies across an image, leveraging light uniformity analyses developed to detect mura (cloudiness) in illuminated displays in order to locate and quantify unprogrammed defects in unpredictable locations. Combining this contextual defect evaluation, human visual perception, and superior imaging quality, a photometry-based vision system blends the benefits of automation with human visual acuity and judgment.

There are five key areas in which imaging photometers provide advantages over standard machine vision and human inspection for quality control:

1. Defects on Parts Are Small, Subtle, or Low-Contrast

Photometry-based cameras offer higher resolution and dynamic range compared to standard machine vision cameras, making them better equipped to detect small and low-contrast defects. Designed for light measurement, imaging photometers must capture extremely detailed images in order to analyze light distribution and color variation at the pixel level. Standard vision systems, on the other hand, are more commonly used to distinguish differences between connected “blobs” of pixels with a baseline uniformity, and there is usually little need for pixel-to-pixel measurement in their typically high-throughput applications.

Resolution:

The CCD resolution of an imaging photometer may be as large as 29 megapixels or, in rare cases, even more. This level of precision can only be replicated by human inspectors on the line, although even subtle details can be missed by the human eye.

Examples of varying resolutions

Dynamic Range:

Dynamic range is the number of grayscale values that can be discerned in an image. A sensor with a broader dynamic range enables a photometric camera to detect hairline variations on surfaces caused by shadows or reflections of light (like shadows that indicate a scratch, or spectral reflections that indicate a metal component).

Three images showing varying dynamic ranges

This, coupled with the high resolution of photometric cameras, enables photometry-based vision systems to image and classify small and low-contrast defects with extreme precision, while maintaining image processing speed and low image noise. (Image noise is further reduced in imaging systems featuring internal cooling of the camera’s CCD). In some cases, a photometric imaging system can detect defects so subtle that they are often unnoticed by human inspectors. These include light scratches on glass, gap variations of less than 1 mm between keys in a keyboard, or the absence of tiny black screws from drill holes in a black surface (as seen in the example below).

Inspecting the presence of black screws on a black surface

2. Defects are Unpredictable and Occur Randomly

The more complex an assembly, the greater the chance an error will occur during production – and, the more difficult it is to know where and when it will occur. Human inspectors have the benefit of applying superior judgment to visual inspection, using context to evaluate whether a defect exists. They do not need to be told where to look for a defect on a part to spot an anomaly, which is beneficial in environments where a number of defects can occur in a given assembly.

A standard machine vision system must be programmed with static points of interest to know where to “look” for defects. This requires advanced knowledge of defect locations and a consistent production environment where parts do not change significantly. For highly-complex assemblies, this pre-programming of the vision system can require hours of work – and new defects may still go undetected. Randomly-occurring errors such as dents, smudges, scratches, and other surface abrasions are the most difficult for a standard machine vision camera to identify and inspect because their location, size, and scope varies so unpredictably.

Illuminated display containing color mura         Photometric detection

Example of an illuminated display containing color mura in the upper-left quadrant (left) and photometric detection (right).

Photometry-based vision systems leverage their light measurement capability to detect and quantify unpredictable assembly and surface-level defects. One example is the use of “just noticeable difference” (JND) evaluation for uniformity measurement. This value represents the amount of change over a surface that enables a difference to be visually noticeable to a standard observer at least half of the time. Many photometric (as well as colorimetric) imagers utilize JND to evaluate significant non-uniformities and mura (cloudiness) in illuminated displays that would be deemed unacceptable by a human observer. JND measurement can be applied to non-illuminated surfaces for detecting and grading “uniformity” issues on surfaces caused by subtle variations in contrast that deviate from expected tolerances, indicating scratches, dents, smudges, and other defects.

Smudges detected on tablet screen using principles of uniformity evaluation.

Smudges detected on tablet screen using principles of uniformity evaluation.

A dent’s length and width can be measured to determine whether a device passes or fails based on the severity or scope of the defect.

A dent’s length and width can be measured to determine whether a device passes or fails based on the severity or scope of the defect.

3. Quality Issues Need to be Quantified and Recorded

Unlike human inspectors, machines can process an unlimited number of data points simultaneously, apply values to quantify each data point, process and communicate data with speed (even wirelessly), and store large amounts of data for long periods of time without losing important details. Most machine vision systems are capable of documenting defects as digital data. However, standard machine vision systems have a limited ability to measure and grade discrete defects to evaluate levels of severity that may warrant part rejection. This is necessary in applications where a margin of error may be acceptable below a given threshold – for instance, logos at a slight degree of misalignment, or dents of a relatively shallow depth. These thresholds may vary within the context of the overall part size, texture, complexity, or other features, making an error more or less obvious to a human observer.

Cable correctly placed = pass       Cable incorrectly placed = fail

Imaging photometers are able to acquire more data from an image than standard machine vision systems, capturing a greater number of pixels and more gray levels to evaluate contrasting points of interest that may indicate a defect, such as the misrouted cable (right).

The difference in capability between standard machine vision systems and photometry-based systems for quantifying defects harkens back to their differences in resolution and dynamic range. An imaging photometer is able to acquire more data from an image than a standard machine vision system, capturing a greater number of pixels and more gray levels to evaluate contrasting points of interest that may indicate defects. For example, at a detectable blob size, a defect can be measured by a standard camera (output in pixel count, length, width, or other values) but standard machine vision optics limit the evaluation of extremely small, incremental measurements or contrast differences.

Alternatively, advanced inspection systems utilizing imaging photometers may offer software tools that assign defects a “visibility score.” This score is calculated by comparing physical dimensions of a defect in detail (less than a millimeter measurement) against programmed tolerances, thereby enabling the evaluation of defect severity within the scope of the imaged part. If a defect’s visibility score is too high, the part may be rejected. In other cases, a defect’s visibility score may fall within an acceptable margin that indicates that a part is able to be repaired and returned to the line. Defect trending data acquired by the inspection system can also be used for pareto analysis, helping to limit errors and reduce the number of rejected parts over time.

Photometric software assigns a “visibility score” to each defect on a part, grading the severity of the defects based on defined tolerances.

Photometric software assigns a “visibility score” to each defect on a part, grading the severity of the defects based on defined tolerances.

Manufacturers can then set pass/fail tolerances to accept or reject defects of a given visibility score.

Manufacturers can then set pass/fail tolerances to accept or reject defects of a given visibility score.

4. Consistency and Repeatability of Inspections Is Important

Humans provide heightened visual acuity and judgment to identify and grade critical defects. However, they make inconsistent determinations from person to person and do not provide repeatable, actionable data to enable traceability like an automated solution does. In addition, they are easily fatigued when studying intricate arrays and configurations of components for extended lengths of time. The accuracy of human inspection decreases precipitously over time, and becomes even worse as the complexity of an assembly increases. Defects on flat products like keyboards and keyboard frames, for example, which exhibit visual patterns ranging from complex arrays to entirely random features, are more likely to be missed by human inspectors who are easily fatigued scrutinizing parts with a high degree of detail. However, catching minute defects in complex arrays may be critical for preventing latent failures in these assemblies.

Human Inspecting a circuit board

Humans provide heightened visual acuity and judgment to identify and grade critical defects. However, they make inconsistent determinations from person to person and do not provide repeatable, actionable data to enable automation.

In cases where absolute defect detection of complex assemblies is required, manufacturers may employ several human inspectors to provide repeat inspections of parts to ensure all errors are detected. This is a costly method, especially on production lines where downtime is limited, requiring several shifts of multiple inspectors on the line for consistent quality control. Alternatively, imaging photometers realize the benefit of continuous operation with consistent, repeatable defect detection performance. Automated systems do not lose their efficacy over time, and are able to apply consistent defect evaluation regardless of part complexity, since performance is not affected by limited data capacity. Because they are programmed with defined tolerances for quality, performance remains consistent over time.

Photometric imaging systems analyze images to quantify defects using precise spatial measurements, which is especially advantageous to manufacturers requiring consistency across multiple lines and factories. These measurements are compared to a standard tolerance value to ensure that inspections from line to line and factory to factory are being evaluated using the same criteria, maintaining quality control consistency and process repeatability.

5. Humans or Other Technology Have Failed to Meet Goals

Perhaps the most compelling reason to evaluate photometry-based imaging for an inspection application is the failure of other solutions to adequately meet the manufacturer’s needs. Some manufacturers continue to rely on human inspectors simply because machine vision and other automated technologies have been unable to detect critical defects or match the accuracy of human judgement to evaluate defect severity. For these manufacturers, the cost of an escaped error into the supply chain is too significant to risk switching to an automated inspection process, despite its obvious benefits.

In applications where human inspectors are required for their visual acuity and judgment, but automation is desired, a photometry-based system may be the ideal solution. A highly specialized solution, photometry-based systems are calibrated to unique measurement environments and configured by to meet the exact criteria of an inspection application. Each system offers a unique suite of features to accommodate the most difficult and critical quality control tasks. With the correct application of equipment in place, a photometry-based vision solution provides immediate return on investment in human inspection environments, especially where repeat inspectors or multiple shifts of inspectors are otherwise required to achieve desired throughput and accuracy goals.

Conclusion

More accurate, consistent, and cost-effective than humans, and more capable of discriminating visual detection and analysis than standard machine vision systems, imaging photometers offer a specialized solution for the most challenging visual inspection applications. Imaging photometers offer resolutions up to 29 megapixels, dynamic range of over 70 dB, with low noise, cooled CCD operation, and scientific calibration for accurate evaluation of light reflected from part surfaces. This optical sensitivity allows imaging photometers to capture the clearest images for automated visual inspection, ensuring that applied software tools can accurately detect and classify assembly and surface defects. Leveraging photometry-based evaluations of just noticeable differences (JND) to evaluate the scope and severity of defects to match human judgement, imaging photometers are also a suitable alternative to human inspectors on the line, even surpassing human visual perception in some cases. With this balance of human and mechanical advantages, manufacturers employing photometry-based systems can realize improved product quality and production efficiency for a clear return on investment.

Photometric imaging solutions bridge the gap between machine vision systems and true human vision, offering consistent, quantifiable data that matches the visual acuity of human inspectors. With this balance of human and mechanical advantages, manufacturers employing photometry-based systems can establish automated operations for the most challenging inspections.

 

Search AIA:


Browse by Products:


Browse by Company Type: