Do you remember this dress picture (Figure 1)? It was posted and went viral in 2015, dividing the world and taking over all our feeds on social media . However, while everyone was debating whether the dress was white and gold, or blue and black, this social experiment made us realise that the rest of the world did not see what we did. While this was only distributed for entertainment, the same effect can lead to serious consequences in industry or agriculture.
In the agriculture and food sector, the visual aspect is an essential part for the quality of most products. Colour is an indicator of the maturity of the fruit, the freshness of the meat or fish, and the correct quality of milk. It comes as no surprise that the majority of the quality processes include a visual inspection or analysis of the colour.
Product visual inspection performed by an operator is a recurrent action due to its simplicity and quickness. But, is it objective enough? Scientific studies  demonstrate that the results of these inspections can be influenced by psychophysical factors like age, sex, observation ability, experience, temperament and creativity. It also depends on the formation of the operator and the scope of the decision or the feedback of the previous inspections. Other influential aspects come from the work environment: light, noise, temperature, etc. Finally, social aspects like team communication, pressure or isolation also factor in.
These variations in the colour inspection can result in an erroneous classification of the products in different qualities, creating conflicts with suppliers and clients, resulting in an incorrect treatment of the product, and what is more important wasting time and money.
Luckily, there are optoelectronic devices that can handle these tasks correctly in a simple manner. For example, a colorimeter is a tool that measures the colour in an objective way. Colour is what human vision is able to detect from the combination of the outputs produced by our three different retina cones. In a simplified way, colour is the combination of red, green, and blue (related to the different types of cones). However, these limited variables (red, green, and blue) might not be enough to solve some problems. In these cases, it is required the utilization of other instruments, such as spectrometers, that permit to get extra information from light.
A spectrometer is a device that goes further and measures the spectrum of light with hundreds or thousands of variables, detecting the response at different wavelengths, from UV going through violet, blue, green, yellow, red, and infrared. So, in addition to a greater number of variables, it is also capable of detecting the response in regions of the spectrum which are not visible to the human eye like ultraviolet or near infrared. In particular, the near infrared is a sensible and well-known region to detect organic parameters.
For example, Figure 2 shows two different spectra obtained from imaginary vegetable products at different maturation stage. Here, the adequate study and processing of the information obtained from the spectra permit to obtain rapid and non-destructive indicators for product maturity, freshness, and many others.
Like the human eye can detect the maturity of a tomato by seeing the colour green or red, the spectrometer with all its data and accuracy provides a more complex correlation that can measure parameters which are impossible to detect with the naked eye, like humidity, fat concentration and fibre content.
By applying statistic algorithms and artificial intelligence methods to the spectrum of a product or raw material, an analysis of a parameter of interest can be conducted. Hence, a large sample with representative measures composed of both the parameter value and the spectrum of the product is required. From this sample, a mathematical model can be implemented, to provide an estimation of the desired parameter with just the spectrum of the new samples whose parameters are unknown.
Some advantages of this technique are:
- Non-destructive detection of the sample.
- No need for previous preparation of the sample.
- Fast results.
- No physical contact with the sample.
- Determination of qualitative and quantitative parameters.
Going back to the picture of the controversial dress at the beginning of this post, it is a clear example that human vision is not objective and can originate inaccurate results and errors, which can severely compromise the quality of the final products. For this reason, it is essential to use the technology for a standardised and rationalised operative decision for our product quality process.
 The dress. Article on Wikipedia about the viral dress picture.
 A. Kujawińska and K. Vogt, “Human factors in visual quality control,” Manag. Prod. Eng. Rev., vol. 6, no. 2, pp. 25–31, 2015.