Development and application of “The Two Image” method for accurate object recognition and color analysis

Color of rainbow trout (Oncorhynchus mykiss) fillets by image and sensory analysis, and correlation with SalmoFan numbers

A total of 72 male and 50 female trout (Oncorhynchus mykiss) fillets were weighed (range 328-794 g, mean 546.2 ± 101.8 g; and range 426-994 g, mean 672.2 ± 106.1 g, respectively), the pictures of whom were taken in a light box, and image analysis was done to measure pixel colors, length, and view area of the fillets. Weight (W) was predicted using view area (V) obtained by image analysis using linear (W = A + BV), and power (W = AV^B ) equations. R² values were between 0.823 and 0.937. Although there was no difference between mean L* and a* values of male and female fillets, significant differences were found between mean b* values (p < 0.05). The colors of SalmoFan™ (SF) mini were also measured by image analysis and their mean L*, a*, b* values, and their entire color index (ECI) and reduced RGB values from 122 images were calculated. A total of 96 untrained panelists were asked to select the SF color of 5 representative fillets and to designate which point on the fillet image best described the SF color chosen. To predict SF numbers of the fillets by image analysis, four cases were considered: (1) whole fillet, (2) whole fillet with pixels a* > 25, (3) a rectangle along the length of the fillet to approximate panelists' selection, and (4) pixels in this rectangle with a* > 25. Mean L*a*b* values, mean reduced RGB values, and mean ECI of the four cases were used to predict fillet SF numbers. Different results obtained imply that image analysis can do repeatable and objective SF color classification of fillets, depending on the pixel selection method, and the color representation. PRACTICAL APPLICATION: Rainbow trout fillets can be assigned SalmoFan (SF) numbers using image analysis of the fillets. However, the selection of pixels and the color representation method affect the results. If these are standardized, SF numbers can be assigned objectively and automatically.

Image analysis-based quantification of the visual attributes of fish, with emphasis on color and visual texture

Images of Red Sea goatfish Parupeneus forsskali were taken in a light box to perform color and visual texture analyses. The average L* and b* values did not change significantly during storage of 7 days, but the a* values decreased (P < 0.05). The change of visual texture parameters energy and entropy (calculated based on histograms, and based on co-occurrence matrices [COM]), box counting-based fractal results, and texture change index (TCI) values are presented. The appearance of fish became “smoother” over time. The entropy values calculated by histograms decreased with storage (P < 0.05), and the maximum range was 0.395. That for COM-based entropies was 71.96. TCI also decreased with storage (P < 0.05) with a maximum range of 10.67. However, energy values increased during storage. The maximum range of the energy values calculated by histograms over time for any color channel was 0.0036. That for COM-based energies was 5.7. There was no observable change in fractal dimension. These image analysis-based parameters were compared with sensory analysis. A trained sensory panel evaluated the visual texture of a sub-set of images. The R2 values for equation fit between sensory score and texture features were, in increasing order: COM based energy (0.185), COM-based entropy (0.313), histogram-based energy (0.375), histogram-based entropy (0.386), TCI values (0.575). Since TCI correlated better with sensory values, it is recommended to be used in this type of visual texture evaluation.

Effects of Fat Polymorphic Transformation and Nonfat Particle Size Distribution on the Surface Changes of Untempered Model Chocolate, Based on Solid Cocoa Mass

This study aims to understand the bloom process in untempered chocolate by investigating the polymorphic transformation of cocoa butter and changes in chocolate surface. Cocoa mass with varying particle size distributions (PSD) were used to produce untempered model chocolate. Optical microscopy showed that during 25 d of storage, the chocolate surface gradually became honeycombed in appearance with dark spots surrounded by white sandy bloom areas. In conjunction with X-ray diffraction this indicates that the polymorphic transformation of form IV cocoa butter to more stable form V crystals caused the observed surface changes with the most significant changes occurring within 6 d. As bloom developed the surface whiteness increased, but the PSD of nonfat particles showed limited impact on the changes in whiteness. Moreover, scanning electron microscopy showed separated fat crystals on fat-rich dark spots and empty spaces between particles in bloom areas suggesting redistribution of fat in the chocolate matrix.

PRACTICAL APPLICATION

The results reported in this work can facilitate the understanding of fat bloom formation in untempered chocolate with respect to the changes in microstructure and surface appearances. It also contributes to show the details of IV-to-V polymorphic transformation in the fat phase as time went by.

Gradual Reduction in Sodium Content in Cooked Ham, with Corresponding Change in Sensorial Properties Measured by Sensory Evaluation and a Multimodal Machine Vision System

The European diet today generally contains too much sodium (Na(+)). A partial substitution of NaCl by KCl has shown to be a promising method for reducing sodium content. The aim of this work was to investigate the sensorial changes of cooked ham with reduced sodium content. Traditional sensorial evaluation and objective multimodal machine vision were used. The salt content in the hams was decreased from 3.4% to 1.4%, and 25% of the Na(+) was replaced by K(+). The salt reduction had highest influence on the sensory attributes salty taste, after taste, tenderness, hardness and color hue. The multimodal machine vision system showed changes in lightness, as a function of reduced salt content. Compared to the reference ham (3.4% salt), a replacement of Na(+)-ions by K(+)-ions of 25% gave no significant changes in WHC, moisture, pH, expressed moisture, the sensory profile attributes or the surface lightness and shininess. A further reduction of salt down to 1.7-1.4% salt, led to a decrease in WHC and an increase in expressible moisture.

Quantification of visual characteristics of whipped cream by image analysis and machine vision: method development

The appearance (color, shine, surface roughness, bumpiness, view area, height) of whipped cream samples change with time. Quantitative methods based on image analysis were developed to measure these parameters in cream samples stored at 3 temperatures (room temperature, 10 °C and 2 °C) for up to 24 h. There was a small decrease in L* and a* values over time, and a significant increase in b* values. ΔE values suggest that these color changes are significant and observable (2.96 for room temperature, 9.46 for 10 °C, 12.6 for 2 °C). Difference between polarized and nonpolarized images was used to quantify shine. The length of a laser line over the cream sample was measured to quantify the change in bumpiness of the surface (33.8%, 41.78%, and 33.27% reduction in laser line length for room temperature, 10 °C, and 2 °C, respectively). For room temperature stored samples, most changes occurred during the first 5 min. For samples stored at 10 °C and 2 °C, most changes occurred during the first 30 min. Turn angles data did not provide useful information. The changes in view area depended on temperature: at room temperature the area increased over time, at 10 °C it increased first then decreased; at 2 °C the view area decreased over time. Correlation of the results of these methods with sensory evaluation can make the evaluation of the appearance of whipped cream more objective, repeatable, and quantitative.

Effects of dry brining, liquid smoking and high-pressure treatment on the physical properties of aquacultured King salmon (Oncorhynchus tshawytscha) during refrigerated storage

BACKGROUND

Aquacultured King salmon (Oncorhynchus tshawytscha) pieces were dry brined with a salt/brown sugar mix, dipped in liquid smoke for 3 min, vacuum packed, high hydrostatic pressure (HHP) treated at 600 or 200 MPa for 5 min and stored at 4 °C for up to 40 days.

RESULTS

The surface redness (average a*) of the samples increased after dry brining, then decreased after liquid smoke treatment. HHP did not change the outside color of liquid-smoked samples. However, the inside color changed depending on pressure. HHP-treated control samples without dry brining and liquid smoking changed to a pale pink color. HHP at 600 MPa resulted in a significant increase in hardness. Compared with fresh samples, dry-brined samples had reduced water activity, while samples dipped in liquid smoke had lower pH values.

CONCLUSION

Dry brining and liquid smoking protect the outside color of salmon against changes caused by HHP. The increase in hardness may counteract the softening of the smoked salmon tissue over time.

Color change of the snapper (Pagrus auratus) and Gurnard (Chelidonichthys kumu) skin and eyes during storage: effect of light polarization and contact with ice

Ten gurnard and 10 snapper were stored on ice. One side always contacted the ice; the other side was always exposed to air. At different intervals for up to 12 d, the fish were placed in a light box, and the images of both sides were taken using polarized and nonpolarized illumination. Image analysis resulted in average L*, a*, and b* values of skin, and average L* values of the eyes. The skin L* value of gurnard changed significantly over time while that of snapper was substantially constant. The a* and b* values of both fish decreased over time. The L* values of eyes were significantly lower for polarized images, and significantly lower for the side of fish exposed to air only. This may be a concern in quality evaluation methods such as QIM. The difference of colors between the polarized and nonpolarized images was calculated to quantify the reflection off the surface of fish. For accurate measurement of surface color and eye color, use of polarized light is recommended.

Proof of concept of using chromogenic arrays as a tool to identify blue cheese varieties

A new chromogenic array for the identification and classification of blue cheeses has been developed. It is based on the response of a chromogenic array composed of five sensing materials prepared by the incorporation of pH indicators to MCM-41 and alumina. Four blue cheeses were tested: Roquefort, Blue Stilton, blue cheese with leaves and blue cheese spread. The colour modulations of the chromogenic array were processed by the principal component analysis (PCA) and partial least squares discriminant analysis (PLS-DA). The statistical PCA analysis showed different responses to each cheese. PLS-DA models were developed by incorporating the data measured at diverse times, and this approach allowed us to obtain a perfect classification of all five cheeses in 5.5h. The results suggest that chromogenic arrays and optoelectronic noses can be a suitable approach to develop simple systems to classify blue cheeses and of potential use for the detection of food fraud.

Effect of frying temperature and time on image characterizations of pellet snacks

The development of non-destructive methods for the evaluation of food properties has important advantages for the food processing industries. So, the aim of this study was to evaluate the effects of frying temperature (150, 170, and 190 °C) and time (0.5, 1.5, 2.5, 3.5 and 4.5 min) on image properties (L*, a* and b*, fractal dimension, correlation, entropy, contrast and homogeneity) of pellet snacks. Textures were computed separately for eight channels (RGB, R, G, B, U, V, H and S). Enhancing the frying time from 0.5 min to 2.5 min increased the fractal dimension; but its increase from 2.5 min to 4.5 min could not expand the samples. Then, the highest volume of pellet snacks was observed at 2.5 min. Features derived from the image texture contained better information than color features. The best result was for U channel which showed that increasing the frying time increased the contrast, entropy and correlation. Developing the frying temperature up to 170 °C decreased contrast, entropy and correlation of images; however these factors were increased when frying temperature was 190 °C. These results were invert for homogeneity.

Image analysis method to quantify the effect of different treatments on the visual meat/shell ratio of half-shelled green lipped mussels (Perna canaliculus)

BACKGROUND

Aquacultured green lipped mussel (Perna canaliculus) is the New Zealand export leader of seafood in terms of weight. Different treatments shrink mussel meat differently and affect the consumer perception of half-shelled mussels. In order to quantify this, digital images of half-shelled green lipped mussels subjected to two postharvest treatments (ultrahigh pressure (UHP) and heat treatment (HT)) and raw controls were taken. The ratio of the view area of the meat to that of the shell (labelled as 'visual condition index' (VCI)) was measured using image analysis. A polygonal region of interest was defined on the image to depict the boundary of the meat and to calculate the view area.

RESULTS

Raw mussels had a VCI of 85%. HT mussels had a much reduced VCI of 41%, indicating shrinkage of the meat due to heat. UHP treatment used as a shucking method resulted in a VCI of 83%. Since VCI is one measure of quality for the consumer, this quantitative method can be used in the optimization of shucking treatment (HT or UHP).

CONCLUSION

VCI can be used to optimize postharvest treatments to minimize meat shrinkage. This method can also be applied to other shellfish such as oysters and clams.