The feasibility of animal source foods’ color measurement using CVS

Color assessment of animal source foods was investigated using a computer vision system (CVS) and a traditional colorimeter. With the same measurement conditions, color readings varied between these two approaches. The color measured by CVS was highly similar to the actual color of animal source foods, and ranged from 75.0%-100.0% of actual colors, whereas colors read by a Minolta colorimeter showed non-typical appearances. The CVS-obtained colors were more similar to the color of food visualized on the monitor, compared to colorimeter-generated color chips. Considering these results, it could be concluded that the CVS is a superior alternative for replacing traditional devices by providing better accuracy.


Introduction
Animal source foods provide numerous essential compounds in human nutrition [1][2][3]. Regardless, color is still the most significant sensory attribute when it comes to consumers' food decisions [4]. It is a vital tool in marketing, especially in food marketing [5].
As far as meat color is considered, darker color is less preferred by customers, who connected it with a lack of quality [6]. This physical attribute can be a measure of some imperfection in milk, such as adulteration [7] or spoilage [8] and can indicate long-term storage conditions [9]. Moreover, consumers desire yellow-orange egg yolk rather than off-white yolk [10][11].
Objective color evaluation is crucial for food technology. Currently, the most common color measurement devices are Minolta colorimeters [12,13]. These devices offer simple and fast food color analysis, and moreover, they are easy to handle and calibrate. Each colorimetric instrument has various settings influencing food color parameters such as color system, illuminant, observer and port size. However, only a few researchers reported all the procedures and technical parameters used for meat and milk color determination, as stated by Tapp et al. [12] and Milovanovic et al. [13]. The majority of papers reported using illuminant D65, 8 mm aperture size and 10° standard observer for instrumental color measurement of milk and meat.
However, the colorimeter has numerous drawbacks regarding inability to capture broad spectral information related to internal characteristics of subjects [14] and a requirement for subjects with homogenous color [15]. In general, these color devices require homogeneous and uniform samples to achieve consistent analysis [16]. To overcome shortcomings of colorimeters, a new, alternative method, the computer vision system (CVS), has been developed. By applying the CVS, the color readings can be determined for each pixel of a sample image, and the technique is rapid, cost-effective and simple [17]. Additionally, CVS has been widely used for color measurement of animal source foods [18][19][20][21][22].

Milk and milk products
The color coordinates of milk and milk products were statistically different when determined by the two devices (CVS and colorimeter) as reported by Milovanovic et al. [22].
In terms of milks, the samples seemed lighter and redder when CVS was used than when the colorimeter was used. On the other hand, all milks had higher b* readings, resulting in more yellow milk appearance, when read by the colorimeter than when acquired by CVS. According to the color difference scale, these two different devices provided easily perceptible total color difference, from 4.3 (cows' milk and goats' milk) to 5.6 (sheep's milk) [22]. The color parameters of raw milks read by colorimeter were in line with the literature color data reported by Milovanovic et al. [13]. Overall, dairy products with a dominant white color read by colorimeter had higher L* (brighter color), lower a* (greener color) and higher b* readings (yellower color) as compared to the CVSacquired colors [22].
The color of white cheeses assessed by colorimeter was lighter than color acquired by CVS. White cheeses were closer to the red and blue region when color was acquired by CVS as compared to the green and yellow region read by the colorimeter. Color differences ranged from 11.3 to 11.8 [22].
As regards liquid fermented dairy products, all L* and b* readings read by colorimeter were higher than by CVS, whereas a* readings were more in the redness region when color was acquired by CVS compared with colorimeter-produced color. The color variations were in line with the color differences and ranged from 5.8 (yoghurt) to 6.6 (kefir) [22].
Color determinations using the two devices for color detection of sour cream and heat processed cream were significantly different. Moreover, using the colorimeter resulted in a brighter, greener and yellower appearance as compared to the color acquired by CVS. The total color differences ranged from 6.7 (heat treated cream) to 11.0 (sour cream) [22].
When it comes to the skim milk powder, there was a significant difference between colorimeter and CVS color readings. L* measured by colorimeter had higher values than L* acquired by CVS. On the contrary, all a* values acquired by CVS were higher (more red) than those measured by the colorimeter. Yellowness values measured by the colorimeter were higher (yellower appearance) than those acquired by the CVS [22].
With regard to the lightness observations of kajmak spread, the colorimeter produced higher values (brighter appearance) than the CVS [22]. All a* values acquired using CVS were less green, in contrast to the colorimeter-measured color, whereas all the b* values indicated a more yellow color with the colorimeter, in comparison to the CVS. The overall color difference was 9.5, indicating the difference in suggested color would be perceptible at a glance.
Dairy products with a dominant yellow color, on color acquisition by the CVS, showed darker (apart from Grana Padano), more red and more blue appearance [22], as compared to the colorimeter. All yellow cheeses, except Grana Padano, on color acquisition by the CVS, showed darker color than was measured with the colorimeter. Regarding a* observations, the CVS resulted in more red appearance, or colors obtained by the colorimeter were less green. The total color difference ranged from 6.0 for pasta filata to 14.9 for processed cheese, indicating large color differences [22].
Regarding butter color, a* values acquired by CVS were higher than those measured by the colorimeter, indicating a less red appearance. In contrast, yellowness data were higher with the colorimeter than by CVS. There was a great total color difference, 11.8 [22].
Apricot fruit yoghurt had different color data as read by the colorimeter and the CVS [22]. Colorimeter-generated color was lighter in terms of L* value. Furthermore, the redness parameter was higher with the CVS than with the colorimeter. Yellowness was higher with the colorimeter than with the CVS, denoting a more yellow appearance of this fruit yoghurt.
The color of whey powder as acquired by the CVS was significantly darker, more red and less yellow compared with colorimeter-measured appearance. The total color difference was 17.1, indicating a large color difference [22].

Eggs
The color parameters of eggshell measured by the two approaches were statistically different with some exceptions (L* reading for quail's eggshell and WI for turkey's eggshell). The color of eggshell gathered through the colorimeter depicts a brighter, less red and more yellow appearance than the CVS-acquired color.
Regarding the color of egg yolk, the colorimeter produced a lighter (except goose egg's yolk), more green and less yellow color, whereas the CVS indicated the appearance of albumen as lighter (except quail's egg albumen), more red and less yellow than the colorimeter.

Conclusion
Even if the same subjects and parameters for color evaluation were studied, significant differences were observed in the color properties measured by the two systems. The colorimeter was less representative and less precise for measuring the color of animal source foods. The reason for this was light penetration, the amount of which related to the device used. In the CVS, the lamps are placed 50 cm above the subject and the light hits the surface and only penetrates a few mm into the subject, whereas the colorimeter is positioned onto the subject surface, and the light penetration through the food matrix must be higher than for CVS. Therefore, the CVS should be seriously taken into account as an effective and more powerful alternative to the colorimeter and as a non-contact tool for measuring the color of animal source foods.