Quantitative Detection Method for Goose and Duck Mixed Feathers by Hyperspectral Camera

In the textile industry, goose down and duck down have become high-quality raw materials for making high-end thermal insulation products due to their excellent thermal insulation properties. However, there is a significant difference in market prices between goose down and duck down. Some unscrupulous merchants, in pursuit of high profits, often mix duck down into goose down to pass off inferior products as high-quality ones. This not only harms the interests of consumers but also disrupts the market order. Therefore, it is particularly important to accurately and efficiently conduct quantitative detection of goose and duck mixed down. In recent years, the development of hyperspectral camera technology has provided an innovative solution to this detection problem.

I. Sample Preparation: Collect a large quantity of pure goose down and duck down samples, ensuring their sources are reliable and representative. Use a high-precision electronic scale to accurately weigh the goose down and duck down in different proportions, and prepare a series of samples with known mixed ratios of goose and duck down, such as setting samples with different duck down mixing ratios of 5%, 10%, 15%,... 95%, each with multiple replicate samples to improve the accuracy and reliability of the experiment. Evenly spread the prepared mixed down samples on a specially designed sample platform, ensuring uniform distribution without overlapping or gaps, to ensure that the hyperspectral camera can obtain comprehensive and accurate spectral information.

II. Hyperspectral Image Acquisition: In this paper, a hyperspectral camera with a spectral range of 400-1000nm was applied. The product FS13 from Hangzhou Cai Pu Technology Co., Ltd. can be used for related research. The spectral range is from 400 to 1000nm, with a wavelength resolution better than 2.5nm and up to 1200 spectral channels. The acquisition speed for the full spectral range can reach 128FPS, and after band selection, the maximum speed is 3300Hz (supporting multi-region band selection). Multiple shots were taken for each mixed fluff sample, and images were obtained from different angles to reduce detection errors caused by local feature differences of the samples. After each shot, the collected hyperspectral image data was promptly transmitted to the computer for storage to avoid data loss.

III. Data Preprocessing: Utilize professional data processing software to conduct preprocessing on the hyperspectral image data collected. Firstly, perform radiometric correction to eliminate the radiometric errors caused by the performance differences of the camera itself and environmental factors, ensuring that the spectral data of different images are comparable. Then, carry out geometric correction to correct the image deformation caused by factors such as the camera's shooting angle and the sample's placement position, ensuring that the position of each pixel point in the image is accurate. Finally, apply noise reduction processing to the image, using filtering algorithms to remove the noise interference in the image, improving the quality and clarity of the image, so as to extract spectral features more accurately in the subsequent steps.

IV. Spectral Feature Extraction: For the preprocessed hyperspectral images, specific algorithms and software tools are utilized to extract spectral features for the goose down and duck down regions respectively. Through the analysis and comparison of a large amount of image data, the specific wavelength range that can significantly distinguish goose down from duck down is determined in the visible light to near-infrared spectral region. At these key wavelengths, the reflectance values of goose down and duck down are carefully measured and recorded to form their respective unique spectral feature datasets. For example, after multiple experiments and analyses, it was found that there are obvious differences in the reflectance curves of goose down and duck down within the wavelength range of 700nm - 800nm, and these differences can serve as important bases for identifying the two types.

V. Model Establishment and Validation: Based on the spectral characteristic data of goose down and duck down extracted, using machine learning or statistical methods, establish a spectral model for quantitative analysis of goose and duck mixed down. Common modeling methods include support vector machines, partial least squares, etc. During the modeling process, a portion of sample data with known mixed ratios is used as the training set to train the model, enabling it to learn the intrinsic relationship between the spectral characteristics of goose down and duck down and the mixed ratio. Another portion of samples not involved in the training is used as the validation set to validate the established model. The high-spectrum image data of the validation set samples is input into the model, and the predicted mixed ratio of goose down and duck down is calculated through the model and compared with the actually known mixed ratio for contrast analysis. By calculating the error between the predicted value and the true value, such as root mean square error, mean absolute error, etc., the accuracy and reliability of the model are evaluated. Based on the validation results, the model is adjusted and optimized, such as adjusting model parameters, adding or reducing feature variables, etc., to improve the performance of the model.

VI. Result Analysis and Evaluation: Summarize and statistically analyze the detection results of all mixed down samples. Calculate the average values, standard deviations, and other statistical indicators of the detection results under different mixing ratios, and evaluate the stability and repeatability of the detection method. Compare and analyze the detection results of the hyperspectral camera with those of traditional detection methods (such as chemical analysis methods), further verifying the accuracy of the hyperspectral camera detection method. Through the analysis of a large amount of experimental data, obtain the key performance indicators such as the error range and detection accuracy of the hyperspectral camera in the quantitative detection of goose and duck mixed down. The experimental results show that this method can quickly and accurately detect the precise proportion of goose down and duck down in the mixed down within a short time, and the detection error can be effectively controlled within a very small range, fully demonstrating its high reliability and practicability.

The application of hyperspectral camera technology has greatly improved the accuracy and efficiency of quantitative detection of goose and duck mixed down. For production enterprises, it can ensure product quality and maintain brand reputation; for regulatory authorities, it provides powerful technical support for cracking down on counterfeit and substandard products in the market, helping to purify the market environment and safeguard the legitimate rights and interests of consumers. With the continuous development and improvement of technology, it is believed that the application of hyperspectral cameras in quantitative detection of goose and duck mixed down and other related fields will be more extensive and in-depth, injecting new vitality into the healthy development of the industry.