Application of hyperspectral imaging technology to the detection of protein content in milk

In the evaluation of dairy nutrition, the protein content is the most important indicator that milk is an essential source of protein absorption in People's Daily life. In recent years, the health of consumers and the development of the dairy industry are closely related to the quality of milk. Therefore, the detection of milk protein content is a very important link. Traditional detection methods consume a long time, waste a lot of human resources, and lead to environmental deterioration. Therefore, it is of great significance to find a faster and more accurate method for detecting milk protein content. Therefore, this paper uses machine learning combined with hyperspectral imaging technology to quantitatively evaluate milk protein content, providing a feasible scheme for milk protein content detection on the market. Specific research work and conclusions are as follows:

一、Experimental materials

We bought seven different brands of pure milk, including Mengniu, New Hope, Yili and Guangming, and stored them in the refrigerator. Protein content is shown in Table 1.

二、Experimental equipment

In this paper, a 400-1000nm hyperspectral camera is used. FS13, a product of Hangzhou Color Spectrum Technology Co., LTD., can be used for related research. The spectral range is 400-1000nm, the wavelength resolution is better than 2.5nm, and up to 1200 spectral channels can be reached. The acquisition speed can reach 128FPS in the full spectrum, and the maximum after band selection is 3300Hz (support multi-region band selection).

三、Experimental setting method

The hyperspectral images of milk samples were collected by using the hyperspectral spectrometer. Samples were collected three times for each type of milk, and then a clear image was selected from ENVI5.3. The collected spectral image had a resolution of 777x1004 pixels. The exposure time of the hyperspectral imager was 10ms, the pixel mixing times were 6, the resolution was 4.8nm, the average interval was 0.8nm, the vertical distance was 30cm, and the acquisition condition was room temperature (23~25°C). The imaging spectrometer and scanning head are installed together during the shooting, and the average spectral data of the milk is derived from the hyperspectral image using the ENVI software."

四、Extraction and preprocessing of hyperspectral data

Extracting hyperspectral reflectance data from hyperspectral images is the basis of traditional machine learning modeling. Generally, the spectral reflectance data of samples is obtained by extracting the average spectral reflectance of all pixels in the region of interest (ROD). In this paper, ENVI software was used to open the corrected hyperspectral image of milk sample, and the pixel near the center of each hyperspectral image was selected as the ROI with the rectangle tool. A total of 30 ROI and 7 hyperspectral images were selected, and 210 ROI were selected. The average spectral reflectance of all pixels in ROI was calculated as the spectral data of the sample, a total of 210 spectral data. The spectral data is saved in ASCI format. The following figure shows the process of extracting ROI.

In this paper, hyperspectral imaging technology combined with machine learning was used to predict milk protein content in order to improve the accuracy of milk protein content prediction. Hyperspectral imaging system was built, hyperspectral images of 7 kinds of milk brands on the market were collected, spectral data were extracted by ENVI software, milk hyperspectral data set was established, and 210 hyperspectral data were extracted finally.

Hyperspectral imaging technology has shown great potential in the field of milk protein content detection, although there are some challenges at this stage, but with the integration of interdisciplinary technology innovation, it will gradually revolutionize the traditional milk detection mode. Through continuous optimization of the technical system and solving practical application problems, hyperspectral imaging will become an indispensable and powerful tool for dairy quality control, help improve the economic and social benefits of the milk industry, and meet the growing demand of consumers for high-quality dairy products.