From Inspection to Perfection: Hyperspectral Imaging in Quality Control

 

From machinery malfunctions and human error to variations in raw materials and contamination, many issues can arise during manufacturing. To deal with these challenges, manufacturers must implement proper quality control with dedicated systems and processes to inspect products and ensure standards are being met.

The specifics of a quality control system vary depending on the product and the tolerances or regulations it must adhere to. Examples include: 

  • Visually monitoring production lines to check for issues or defects. Typically, this process is now automated, relying on computer vision systems and machine learning models trained on a specific product or packaging.
  • Quantifying cleanliness to ensure products meet the required standards and function effectively. Key techniques include particle counting and surface cleanliness tests to identify contaminants invisible to the naked eye.
  • Inspecting raw materials and guaranteeing they meet the required standards. Manufacturers often audit vendors across their supply chain and perform batch sampling to check materials against specifications.
  • Live monitoring of production processes such as mixing, cooking, casting, and others. Tracking these processes allows manufacturers to identify and rectify potential problems in real time.

While quality control changes depending on the needs of the product, what stays the same is the need for high throughput and accuracy without spiralling costs. For businesses looking to improve on existing systems, hyperspectral imaging now offers flexible quality control solutions across a range of use cases. With advances in data capture technology and a maturing development environment, new hyperspectral imaging cameras deliver real-time spectral information to enable fast and accurate quality control systems.

What is Hyperspectral Imaging? 

Traditional cameras capture light using three broad channels: red, green, and blue (RGB). While this dataset enables the creation of simple colour images, mimicking how the human eye sees the world, it fails to capture detailed information on the wavelength of light hitting the sensor.

In contrast, hyperspectral cameras use many more channels to capture wavelength data and reveal new information within the image. This means the technology returns both spectral and spatial information, recreating spectra throughout the image. By analysing these spectra, it is possible to identify signatures related to materials and processes present in the image. 

The Benefits of Hyperspectral Imaging in Quality Control

Through spectral analysis and looking at light in more detail, hyperspectral imaging applications have taken root across many industries, including quality control, where it offers a range of potential benefits: 

  • Non-Destructive: Hyperspectral imaging simply measures reflected light (primarily in the visible and near-infra-red region), it doesn’t alter or damage the product in any way.
  • Material Identification: Differentiate between materials based on chemical composition rather than less accurate visual inspection or slower analysis techniques.
  • Defect Detection: Identify minor defects or variations that would go undetected using standard imaging.
  • Flexibility: With use cases across a range of manufacturing sectors and production processes, hyperspectral analysis can be tailored to focus on specific wavelength regions or spectral indices depending on quality control needs.
  • Integration: Hyperspectral data can be integrated into existing industrial machine vision systems to enhance operations and add an additional layer of detection.

Redefining the Trade-Off Between Frame Rate and Spectral Resolution 

In the past, slower frame rates have limited the throughput of hyperspectral-based quality control systems. Capturing and processing the large datasets present in every hyperspectral image would take significantly longer than traditional imaging. But, with advances in snapshot data capture, it is now possible to deploy hyperspectral imaging industrial inspection systems with real-time response.

Using faster, snapshot hyperspectral techniques over line-scanned systems traditionally comes at the expense of spectral resolution. This trade-off reduces the number of wavelength channels in the final dataset, preventing cameras from identifying all the potential intricacies of the materials and processes present.

The Living Optics camera throws away tradition to change how the industry views this trade-off, delivering video-rate hyperspectral data (frame rates of up to 30Hz) while retaining high spectral resolution (96 channels across 440nm-900nm).

By changing the paradigm of what is possible with hyperspectral technology, the Living Optics camera has the potential to deliver next-generation quality control systems for a range of different industries. Some of the most exciting sectors ripe for innovation include analysis across food production and packaging processes and conveyor belt scanning to ensure quality.

Hyperspectral Imaging Quality Control in Food Production and Packaging Lines

Hyperspectral imaging has several existing and potential use cases in food production and packing lines. These include:

  • Inspecting Complex Products in Human Operator Areas

Human operators inspect items for quality in many food production environments, like butcheries and meat storage facilities. Hyperspectral data can distinguish between products more accurately based on varying textures, colours, and detailed spectral data. For example, this allows it to accurately differentiate between lean or fatty meats or ensure uniformity in meat cuts. Finally, it can detect foreign materials and contamination that might be missed by manual inspection. These use cases allow manufacturers to more accurately assign pricing based on quality while reducing food waste and the chances of selling contaminated products.

  • Quantifying Fat Content

Hyperspectral analysis can measure the fat content of individual products non-destructively by identifying unique spectral signatures. This is particularly valuable for processing meats and dairy products, where precise fat quantification is critical for meeting regulatory standards and maintaining product quality.

  • Detecting Leaking Products

Leaking food products pose significant hygiene and safety risks. This problem is exacerbated by the trend towards more eco-friendly packaging, such as cardboard. Hyperspectral imaging can identify even small amounts of leakage by analysing changes in the spectral properties of the packaging material. By autonomously detecting leaks early, downstream cross-product contamination and replacement costs can be avoided.

  • Detecting Heat Seal Integrity in Food Packaging

Heat sealing is vital for preserving food freshness and preventing contamination. Hyperspectral imaging can detect potential heat seal defects that could allow air and bacteria to come into contact with the food product. By identifying these issues instantly in real-time, hyperspectral systems enhance packaging reliability and minimise recalls.

Using the Living Optics Dev Kit in Belt Scanning Mode  

The Living Optics Dev Kit contains a snapshot hyperspectral imaging camera which is designed for use as a hyperspectral video camera  – capturing both spatial and spectral information in a single frame. Although this is ideal for dynamic video scenes such as autonomous driving or handheld inspection it is also possible to use the system in a more traditional scanning mode.

For every video frame, the Living Optics camera captures one RGB image and one array of radiance spectra from over 4000 points across the scene. This enables live data processing of spatial and spectral features from across the image space at 30Hz. Due to the sparse nature of the spectral sampling, it is possible to aggregate multiple frames together into a single, higher resolution, map of the target scene. This is analogous to how you might use a scanning slit spectral imaging system with a motorized scanning system or conveyor belt, but rather than have data from a single line of points the points are distributed across the scene.

Living optics sampling
Comparison of the Living Optics camera’s single-shot spatial-spectral sampling pattern and traditional line scanning techniques. One blue point indicates one spectral sample on the Living Optics Dev Kit. The single shot spatial sampling of a push-broom sensor is indicated in red for reference.

Although the Living Optics camera can match the operation mode of a scanning system by tracking objects and aggregating spectral data for those objects, an additional advantage is the ability to monitor changes in the sample as it moves. This is useful when dynamic processes are occurring along the direction of motion, for example when:

  • Imaging rotating fruit to check for defects over its entire surface on a roller conveyor
  • Materials are tumbling laterally in the field
  • An object needs to be imaged at a range of orientations to compensate for specular lighting glints.

The Living Optics development kit enables a range of applications across handheld and dynamic inspection. But it can also be used in conveyor belt scanning systems, simplifying the transition between laboratory and batch-based experimentation and production line deployment.

In Conclusion 

New, high-framerate hyperspectral imaging technology offers significant advantages for real-time quality control in manufacturing. With non-destructive spectral analysis of production processes, it is possible to more accurately identify potential problems before their consequences become more severe. This enables businesses to reduce the financial, safety, and reputational damage caused by manufacturing issues or poor-quality products.

The Living Optics camera and development kit are the ideal way for developers to investigate the capabilities of this new technology. You can use the same field portable camera across applications, moving the system between different production lines or locations. Plus, the snapshot data capture provides video-rate hyperspectral images to keep up with moving or rolling products on production lines.

With simple integration tools you can start acquiring data quickly and even combine its output with any existing RGB machine learning model alongside new hyperspectral algorithms for enhanced quality control in manufacturing.

Get in touch with our sales team to start building next-generation quality control systems for your use case.

We would love
to hear from you