Chicken Meat and Beef Identification in UV-Vis Absorbance Spectrum by Applying Savitzky-Golay Method

Spectroscopy is one of the techniques suitable for validation of the quality, safety, and hygiene of meat due to some advantages, such as rapid detection capability, ease of use, and non-destructive measurement. Utilizing an ultraviolet-visible (UV-Vis) spectrometer, with a wavelength range of 200 nm to 1100 nm, this paper proposes a method for processing spectrum data profiling of chicken meat and beef. A 360-time reflectance spectrum was acquired from each chicken breast and beef loin at the optimum integration time of 60 ms. Subsequently, the spectrum was converted into an absorbance spectrum, and the apparent noise was excluded by applying the Savitzky-Golay (SG) method. The result shows the difference in spectrum between chicken meat and beef in which the chicken meat has two peaks at the wavelength of 284.79 nm and 422.69 nm, meanwhile, the beef shows three peaks at wavelengths of 282.46 nm, 419.03 nm, and 577.97 nm. It was found that the SG method enhanced the absorbance spectrum, showing the difference in spectrum behaviors in the wavelength range of 450∼600 nm. Consequently, to shorten the computational time, building a light, cheap, and compact prototype reading device within this specific wavelength range will be the next target in the near future.


Introduction
One of the important sources of proteins, vitamins, and minerals for humans is meat and its derivative products.Meat is also an essential element in diet which is highly valued by consumers.Currently, customers have more awareness of consumed foodstuff, thus making food manufacturers get into competition to fulfill customer satisfaction to produce high quality, safe, and hygienic products.Numerous meat validation techniques have been used to ensure meat product quality, but the techniques that are widely used are mostly destructive, time-consuming, and unsuitable for in-situ, rapid, and real-time report utilization.Spectroscopy can be an alternative technique for assessment since it has the advantage of speed, ease of use, and non-destructive measurement [1][2][3][4].
With the advances in the technology of mixing various types of meat, the development of meat detection systems in portable, cheap, light, easy-to-use, and reliable devices is crucial nowadays.Meat spectroscopy is a technique that uses electromagnetic radiation to analyze the chemical and physical properties of meat products for various purposes, such as quality control, safety assessment, authentication, and shelf-life prediction.Recently, there was meat adulteration detection using a spectroscopy-based sensor directed at minced meat [5].Beef with bovine offal and pork with chicken both in fresh and frozen-thawed samples were their experimental targets and visible and fluorescence (Fluo) spectra and multispectral (MSI) acquisition were utilized [6].Meanwhile, spectrometric measurement for meat speciation combined with chemometric methods and its application to determine halal meat species from pork for halal certification was conducted by utilizing Vis-Near Infrared (NIR) and NIR spectrometers with wavelength ranges of 400~1000 nm and 900~1700 nm, respectively [7].Experiments employing the spectroscopy method for halal meat detection, comparing five brands of handheld NIR devices showed all of the sensors could detect the meat products including various types of mixed meat were reported [8].Other research groups have reported that using spectroscopy at a long wavelength NIR (1000~2500 nm) to detect pork adulteration in halal meatballs has the advantages of being fast, easy sample preparation, and low cost [9].
The reflectance measurement methods are usually used to measure the relative proportion of myoglobin forms at meat surfaces [10].Meanwhile, absorbance measurement methods are utilized as well to identify the proportion of myoglobin forms of muscle samples in solution [11,12] and to analyze chemometrics [13].Furthermore, a research report on pork meat freshness identification through its chromophores concentration by converting the reflectance spectrum into an absorbance spectrum and the SG method using a Vis-NIR spectrometer was published [14].In this paper, a UV-Vis spectrometer with wavelengths ranging from 200 nm~1100 nm to detect chicken meat and beef reflectance spectrum converted into absorbance spectrum is utilized.Furthermore, the SG method is implemented to remove apparent noise and the result is presented.In addition, this research also aims to build a database of spectrum characterization of various meats and mixtures as a basis for prototype reading devices to help the advocacy team of halal products in Indonesia.

Experimental Setup and Data Processing
In this experiment, fresh chicken meat and beef samples were obtained from the traditional market in South Tangerang, Indonesia.The samples were chicken breast (Gallus domesticus) and beef loin (Bos indicus).These parts are selected for their rich myoglobin contents compared with other cuts of meat.The samples were fresh and purchased less than 24 hours after the slaughtering process and had not yet undergone any preservation process e.g., being frozen or chilled, to avoid additional water content to the sample, which could reduce the myoglobin content due to dissolution [15].During transportation, the sample was also kept in an air-tight seal bag to prevent oxidation due to ambient air.The dried surface of the sample approximately 1 mm was removed before measurement, and the meat was sliced into ±25x35 mm 2 with a thickness of not less than 5 mm as shown in Figure 1 (a).Any excess liquid on the meat surface was removed, and the sample was placed on a glass substrate as shown in Figure 1 (b) and (c).The reflectance measurement was carried out immediately after the preparation steps were completed.The above-mentioned treatment was also applied to other samples.The experimental setup is depicted in Figure 2. The optical measurement probe is facing downward to the sample.The Ocean Insight DH-2000-BAL incorporating powerful deuterium and halogen lamps as broadband light sources were utilized.A backscatter/reflection with a solarization-resistance probe is used due to the high UV radiation of the light source and to reduce the probe's transmitting ability [16].The probe consists of a 6-around-1 fiber bundle design, with a 6-fiber leg connecting to the light source and a single-fiber leg connecting to a spectrometer, and each fiber has a 400 µm fiber core diameter.The reflected light was then measured by Ocean Insight spectrometer MAYA 2000 Pro.The probe was mounted on the reflection stage with the base covered with black velvet to reduce possible stray light.The distance between the probe tip and the sample was maintained constant at about 5 mm with the diameter of measurement area of ±2.5 mm.Before measuring the sample, a polytetrafluoroethylene (PTFE)-based white reflectance standard was used for the reference calibration due to its capability to reflect 98% from 250~1500 nm [17].A warming-up time of at least 20 minutes was given to stabilize the light source.The sample was measured at 12 different measurement points at the ambient room temperature with the integration time applied of 60 ms to gain reflectance data.One point of measurement was repeatedly captured five times to avoid measurement error, producing five sets of reflectance data and thus 360 data in total were obtained for each chicken breast and beef loin measurement within wavelengths range of 200~1000 nm.The collected reflectance spectrum was then converted into an absorbance spectrum by utilizing equation (1).Afterward, the Savitzky-Golay (SG) method was utilized to reduce apparent noise in the absorbance data.A more detailed explanation of this method will be published separately.
Furthermore, it was normalized and averaged for further analysis.The mean value of all five data for each wavelength point was calculated and the output was used as the representation of the sample point.Subsequently, MinMax scaling was applied to reduce the effect of outliers while highlighting important points from the data.Thus, the data was now in the range of 0 to 1.

Result and Discussions
The typical absorbance spectrum is shown in Figure 3.The apparent noise is obviously seen in the absorbance spectrum raw data before optimization, starting around 450 nm in Figure 3(a).Subsequently, the SG method was utilized to filter apparent noise in Figure 3(a) since this method is much better than the moving average method, which tends to attenuate data, in preserving data features such as peak height and width which are important information in our research [18].Figure 3(b) shows Figure 3(a) after implementing the SG method.The result shows a smooth absorbance spectrum with no noise.It can be seen clearly that the SG method is quite suitable to be implemented in our research for further absorbance spectrum analysis.4(a) shows the absorbance spectra of chicken meat within the wavelength range of 200~1100 nm after all the reflectance spectra conversion by equation (1).Three peaks are always present in every data within the UV and visible light spectrum region.After the third peak, the same behavior happens for all data in which they descend until wavelength of 500 nm.Eventually, the data trend gradually rises along the spectrum up to 1100 nm.Furthermore, all the data for both meat types were normalized and averaged for every wavelength point.After the process, the distinctive comparison of the absorbance spectrum between chicken meat and beef can be distinguished as shown in Figure 5.For the chicken meat spectrum (red line), there are two peaks at 284.79 nm and 422.69 nm.The beef spectrum (blue line) has three peaks at 282.46 nm, 419.03 nm, and 577.97 nm.The difference in spectrum behavior between chicken meat and beef in the wavelength range of 450~600 nm is very important for the development of our prototype reading device marked by a green dashed rectangular.By focusing the scanning process on the specific area, the sweeping time can be shortened more efficiently than sweeping the entire range of wavelengths.

Conclusion
Detection of chicken meat and beef using a UV-Vis spectrometer with wavelengths ranging from 200 nm~1100 nm is presented.The reflectance data from the experiment was converted into absorbance data, and subsequently, the apparent noise was removed by applying the SG method for further analysis.It was found that the spectrum of chicken meat has two peaks at 284.79 nm and 422.69 nm, meanwhile, the beef spectrum has three peaks at 282.46 nm, 419.03 nm, and 577.97 nm.Another more obvious difference between chicken meat and beef spectra is their behavior within the range of 450 to 600 nm.Thus, for further development, these wavelengths will be utilized to build a prototype reading device that must be light, cheap, compact, and applicable for in-situ measurement.

Figure 1 .
Figure 1.Samples preparation: (a) meat processing (b) a group sample of chicken breasts and (c) beef samples

Figure 2 .
Figure 2. Experimental setup for reflectance spectrum measurement

Figure 3 .
Figure 3. Absorbance spectrum of (a) raw data and (b) improved data using the SG method

Figure
Figure4(a) shows the absorbance spectra of chicken meat within the wavelength range of 200~1100 nm after all the reflectance spectra conversion by equation(1).Three peaks are always present in every data within the UV and visible light spectrum region.After the third peak, the same behavior happens for all data in which they descend until wavelength of 500 nm.Eventually, the data trend gradually rises along the spectrum up to 1100 nm.

Figure 4 .
Figure 4.The absorbance spectra for (a) chicken meat samples, and (b) beef samples optimized with the SG method Figure 4(b) exhibits the spectra of the processed beef data.Compared to the chicken meat spectrum, in the 450 nm~600 nm spectrum range, the data is easily distinguishable by five peaks in the UV and visible light spectrum region.After the fifth peak, they decrease significantly from 600 nm and then flat rippling up to 1100 nm.

Figure 5 .
Figure 5.The optimized absorbance spectrum difference between chicken meat (red) and beef (blue)