Applications of the hyperspectral cameras that are revolutionizing the industry

Have you ever wondered how, through hyperspectral cameras, you can make that invisible become evident? Hyperspectral cameras can reveal hidden details in a wide range of sectors. Today, we will focus on learning about some of the main applications from which the most cutting-edge industries already benefit and how yours can do it too.

What hyperspectral cameras are?

Hyperspectral cameras are advanced imaging devices designed to capture information far beyond what the human eye can perceive. How do they do that? These cameras divide the electromagnetic spectre into a multitude of narrow bands, allowing the detection of spectral details and differences which are imperceptible to traditional cameras.

Hyperspectral cameras break down light into hundreds or even thousands of spectral “bands” corresponding to different wavelengths. As a result, they can detect specific information about how the light interacts with objects and materials. Each material has a unique spectral signature that allows its characterization. Using a hyperspectral camera, it is possible to identify with precision what is being seen without contact or destruction of the sample, even if two objects look the same to the naked eye.

Having more information than traditional vision systems, the computer vision algorithms that we make at ATRIA, process and digest the large amount of data collected by hyperspectral cameras. This allows us to unblock a new level of precision in materials identification, anomaly detection and process optimization across various industries.

Hyperspectral cameras vs traditional cameras

Traditional cameras are focused on capturing images in the visible light range. In fact, captured information is processed to imitate as much as possible the human visual perception and obtain realistic images. However, they lack the capacity of the wide spectral information captured by a hyperspectral camera.

A simple way to visualize these camera’s potential is the next: Traditional cameras capture 3 colours (RGB). We can imagine this as if we were taking three different photographs, one in red, another in green and another in blue, and then overlapping them to create a full-colour blue. This is an analogy with which we are all familiar.

Now, consider that a hyperspectral camera, instead of three “photos”, it takes hundreds of them (one for each wavelength), creating a set of data that includes a wide range of information.

This capacity to capture information at numerous wavelengths is what makes hyperspectral cameras so powerful in the precise identification of materials and the detection of subtle variations in the environment, although to the human eye, it may be indistinguishable.

Hyperspectral cameras vs other spectroscopic system

Of course, there are other devices that, in the same way, hyperspectral cameras, make use of the spectroscopic characteristics from light-matter interaction. However, hyperspectral cameras are different in their capability to capture spectral information in a visual context.  Unlike spectrometers, which analyse light intensity at a specific point or a carefully prepared sample, hyperspectral cameras go a step further by generating two-dimensional images. This implies that they have the ability to identify materials and characteristics in a heterogeneous scene.

Spectroscopy systems are valuable tools in scientific research and forensic engineering. However, they are typically used in a controlled laboratory environment due to their sensitivity to environmental conditions. Factors such as vibrations, temperature fluctuations, humidity, or lightning can significantly affect the accuracy of their measurements. Furthermore, the preparation of the samples for their measurement usually requires manual preparation by laboratory technicians. All of this means that, despite being items generally more precise, these systems are not suitable for implementation in industrial environments.

In contrast, hyperspectral cameras have adapted to endure more adverse conditions. Its robustness to operate in less controlled conditions has contributed significantly to its adoption in a variety of industrial applications. Its ability to take pictures from a distance allows its implementation to take measurements directly in automated industrial processes without the necessity to prepare samples or destroy them during the measurement process.

Applications and examples

We explain you some of the most important applications of hyperspectral cameras in different industries and sectors:

• Precision agriculture: hyperspectral cameras are used to evaluate crop health, typically in aerial view mounted on drones. They can detect early signs of water stress, nutritional deficiencies and diseases in plants, which allows farmers to take preventive measures and optimize the use of resources such as water and fertilizers.
• Food industry: in the food industry, hyperspectral cameras are used to overcome strict quality controls of products such as fruits, vegetables and meat. They can detect imperfections, such as dents, decay or the presence of contaminants, guaranteeing that only high-quality products reach the market.
• Chemical and pharmaceutic industry: in laboratories and chemical manufacturing processes, hyperspectral cameras are used to analyse and monitor the chemical composition of the products. Not only it is being guaranteed the presence of the desired chemistry products, but also their concentration, without requiring more invasive methods. This assures the quality and consistency of chemistry products and helps to detect deviations and not desired impurities.
• Recycling: in recycling facilities, hyperspectral cameras play an essential role in identifying and separating materials such as diverse types of plastics, glass and metals based on their unique composition. This approach of precise classification facilitates residue management by allowing a minute separation of the materials that is not possible to achieve with traditional methods.
• Failures and pollutants detection: early detection of failures and pollutants in the processes of manufacturing and inspection of products. These cameras identify pollutants, such as foreign particles or impurities in any production process. Early detection not only guarantees product quality but also allows immediate action to be taken, avoiding unnecessary operations and optimizing the production chain to obtain a higher added value.

Steps to carry out a project with hyperspectral cameras

To develop any project, it is essential to follow a systematic approach to guarantee success. Here, we present some key steps that we follow in ATRIA for the implementation of hyperspectral cameras:

• Identify the application and the objectives: You have to start by defining which is the project proposal and which are the specific objectives. Which is the problem that is being tried to resolve or the question that is being tried to answer? What spectroscopy information is needed to collect? What is the expected benefit after a successful implementation?
• Cost and technology evaluation: One of the more critical aspects to consider is cost. At ATRIA, we carefully study your problem to determine which is the needed technology. It is necessary to determine the spectral signature of the materials to be studied to use a hyperspectral camera of the appropriate spectral range. For many applications, it is enough with a hyperspectral camera of lower performance (¡or even a conventional one!). Avoiding over-engineering and evaluating which is the optimal technology decreases the cost as well as the complexity of the project.
• Development and analysis: Another fundamental aspect is the development of predictive models. These models are based on collected data and they can be applied to predict and identify future behaviours. In a lot of applications, autonomous learning (machine learning) performs a fundamental role. Machine learning algorithms can be used to train models that can recognize patrons and make decisions based on hyperspectral data. Atria develops the most accurate models for every situation, following an innovative philosophy that adapts the solutions and the technic-scientific existent knowledge. Many times, it is not necessary to reinvent the wheel, but it is crucial to customise it according to the specific necessities of each client.
• Implementation of the system in an industrial environment: Once the predictive model has been developed, refined and validated, it is time to apply it in practice. The successful integration of the system in an industrial environment involves its challenges, such as the installation of the equipment in the production line, as well as communication and synchronisation between them. During this phase, specific challenges related to hostile environmental conditions must be addressed, such as image acquisition speed or standardization of the time of exposition according to ambient light. The key to a successful implementation is to ensure that the system works robustly also in real life. In the last instance, hyperspectral cameras in an industrial ambient can lead to significative improvements in quality, efficiency and decision-making in applications variety.

Do you want to apply Hyperspectral cameras to some of your projects? Contact us!