The Challanges of Vision in Food Industry

Hamamatsu presents its solutions of artificial vision, that allow to reduce errors in the selection and packaging of food in an automated way. Therefore, companies can keep up with the growing demand for high-quality products

The demand for high quality food, that satisfies both our food needs and the desire to eat healthy, is constantly growing. Professionals from agricultural field and packaging industry know the importance lying around the inspection of food products during the packaging process. Challenges are complex, such as the need to detect agricultural products contaminated with foreign bodies (stones, debris etc), damaged packaging, wrongly labelled products or spoiled food.

Technology can provide much help through SWIR (Short Wavelength InfraRed) and VIS imaging, which respectively use InGaAs as well as CCD/CMOS sensors, and through X-ray analysis, which bases on light sources and detectors, suited to be employed in processing or production lines. Thanks to specific cameras and sensors, it is possible to create automated devices for the inspection, sorting and quality control of food. These optical systems allow to identify any non-conformity in food products or their packaging.

An example from coffee processing

For example, during coffee processing, beans are scattered on the ground on apposite sheets made of a fabric that specifically allows drying. After a given time they are then raked and poured into a hopper. This procedure, however, does not allow the elimination of foreign bodies such as stones or gravel which, if packaged together with coffee, could create sanitation problems. In this case, machine vision can be employed to detect objects of similar shape and size, such as coffee beans and gravel. In fact, cameras and detectors can be positioned in strategic points along the production line to identify materials based on their SWIR spectra.

Always optimize the selection and inspection process

With a world population currently standing at 7.7 billion, and projected to grow quickly in the next decades, food companies must keep up with the growing demand for high-quality food. It is therefore necessary to constantly optimize the food selection and inspection process. In the past, the food that went through conveyor belts was sorted manually. Today, manual sorting is no longer practicable due to the continuous increase in volumes, which would imply both cost and accuracy problems. Particularly, human error is always lurking because the operator could be misled by similar colors and shapes, and visual inspection alone does not allow to identify defects inside the packaging or below the surface of the product. Moreover, the bigger volume of products travelling along the conveyor belt, and consequently the higher speed, reduce the accuracy of the operations. Lastly, recalling defective products already placed on the market means additional costs.

The answer provided by machine vision

Machine vision enables companies to reduce errors in the selection and packaging of food in a completely automated way, detecting defects that would be difficult, or even impossible, for a human to identify. To optimize the inspection and sorting process, food products can be scanned using high-speed imaging devices such as cameras with InGaAs or CMOS technology while travelling along the conveyor belt. InGaAs cameras are able to detect defects that are situated under the surface of the product and which are invisible to the human eye, while CMOS cameras allow the detection of visible defects.

Controls in the visible and beyond the surface

Hamamatsu has CMOS linear image sensors with spectral range extended from UV to the near infrared (up to 1,100 nm) with different size (from 128 to 4,096 pixels), sensitivity and reading speed. Two-dimensional CMOS image sensors are also available in various configurations, from 30×30 to 1,280×1,024 pixels. An emerging technique in the field of food inspection is hyperspectral imaging SWIR, above mentioned, which allows both spatial and spectral information to be collected at the same time. By exploiting this technique, it is possible not only to detect foreign bodies and invisible-to-eye defects, but also to identify ingredients and classify foods. For example, it is possible to detect the water content inside fruit and vegetables, which is a much interesting point because the moisture content can significantly influence the quality of agricultural products and the duration of their preservation, since an excess of humidity can cause the development of molds. SWIR imaging can also detect the presence of dark spots on the peel of fruit and vegetables, which is an indication that the product has suffered trauma which allowed oxygen to penetrate below the surface, initiating oxidative processes. Although the consumption of bruised or overripe fruit does not involve health risks, it is not pleasant from an organoleptic and aesthetic point of view.

Identification of bruised fruit.
Identification of bruised fruit.

Sensors and cameras for SWIR imaging

SWIR imaging is performed using InGaAs image sensors, that are able to detect the wavelengths in the near infrared region and are featured by high frame rate, low reading noise and high sensitivity. There are two different types of InGaAs image sensors: linear and area. Linear ones are suitable for inline sorting of agricultural products due to their scanning speed and high sensitivity. There are sensors which are sensitive to various wavelengths, with different number of pixels and scanning speed. Most of them have a metal or ceramic package, but cheaper package options are also available. Hamamatsu Photonics provides a complete product range covering various spectral ranges (extended up to 1.7 or 2.5 um), providing both single sensors and modular solutions that can be easily integrated. Area sensors are two-dimensional detectors consisting of a matrix of InGaAs photodiodes and an integrated readout circuit (ROIC). They can be used in various applications including hyperspectral imaging and packaging inspection. Hamamatsu offers several 2D sensors, sensitive up to 2.55 um, with high sensitivity and able to reach high speeds up to 507 frames/s. Finally, InGaAs cameras are a simple plug-and-play solution based on InGaAs 1D or 2D image sensors. They can also be used for the construction of multispectral/hyperspectral (HSI) imaging cameras.

Among coffee beans, it is necessary to eliminate foreign bodies.
Among coffee beans, it is necessary to eliminate foreign bodies.

Identification of packaging defects

Even after packaging, CCD/CMOS and InGaAs cameras can support in the identification of defects in products to be rejected. For example, the former allows the detection of any visible faults, such as dents or punctures, while the latter are suited to identifying leaks of material from the packaging or filling defects of the containers. Both types of cameras are equipped with high acquisition speed and high resolution, essential to analyse goods travelling fast on the conveyor belt.