Hyperspectral technology was used to detect the moisture content of stored rice. This experiment takes 120 rice samples as the research object, collects hyperspectral images of all rice samples, and uses the preprocessing method of multivariate scattering correction to perform noise reduction processing on the original spectral data of the rice samples. Due to the large amount and strong redundancy of the original hyperspectral data, the stepwise linear regression analysis method was used to extract features from the preprocessed data. Finally, a BP neural network quantitative rice moisture detection model was established. Since the modeling effect did not meet the expected goals, a genetic algorithm (GA) and a mind evolutionary algorithm (MEA) were introduced to optimize the weight sum of the BP neural network. threshold. Comparing three rice moisture prediction models: BP, GA-BP, and MEA-BP, the prediction set determination coefficients of the three models all reached more than 0.86. Among them, the MEA-BP model had the best prediction effect, and the prediction set determination coefficient reached 0.966. 3, and the root mean square error is 0.81%

Abstract:

This study investigates the use of hyperspectral imaging technology to accurately measure the moisture content in stored rice, addressing the challenges of quality control in agricultural storage and processing. The experiment encompasses the analysis of 120 rice samples using state-of-the-art hyperspectral imaging combined with sophisticated data processing techniques to enhance the accuracy of moisture content detection.

Introduction:

In agricultural industries, particularly in rice processing and storage, the accurate measurement of moisture content is critical for maintaining quality and preventing spoilage. Traditional methods often fall short in providing real-time and precise moisture content data, necessitating the adoption of advanced technologies such as hyperspectral imaging.

Methodology:

The research utilized hyperspectral imaging to collect detailed spectral data from 120 rice samples. Each sample’s original spectral data underwent noise reduction through multivariate scattering correction—a preprocessing technique that enhances the clarity and usability of the hyperspectral data by minimizing spectral noise and correcting scatter effects.

To manage the voluminous and redundant data typical of hyperspectral imaging, stepwise linear regression analysis was employed. This method streamlined the feature extraction process, focusing on the most relevant spectral features for moisture detection.

Model Development:

The initial phase of modeling involved setting up a Back Propagation (BP) neural network. However, the initial results from the BP model were suboptimal in predicting rice moisture content accurately. To refine the model’s predictive capability, two advanced algorithms were introduced: a Genetic Algorithm (GA) and a Mind Evolutionary Algorithm (MEA). These algorithms optimized the neural network’s weight sum and threshold, significantly enhancing model performance.

Results:

Three distinct models were developed and compared: the standard BP model, the GA-enhanced GA-BP model, and the MEA-enhanced MEA-BP model. Performance metrics included the determination coefficients of the prediction sets and the root mean square error (RMSE) of moisture content predictions.

The determination coefficients for all three models exceeded 0.86, indicating a high level of accuracy in moisture prediction. Notably, the MEA-BP model outperformed the others, achieving a determination coefficient of 0.966 and an RMSE of 0.81%, demonstrating superior predictive precision.

Discussion:

The integration of GA and MEA optimization algorithms with the BP neural network marks a significant advancement in hyperspectral data analysis for agricultural applications. The MEA-BP model, in particular, showcases the potential of evolutionary algorithms to enhance neural network performance in complex predictive tasks.

Conclusion:

This study confirms the efficacy of hyperspectral imaging combined with advanced algorithmic optimization for improving the accuracy of moisture content detection in stored rice. The successful application of these technologies not only enhances the quality control processes in agriculture but also opens avenues for further research into other hyperspectral imaging applications in various industrial sectors.

Future Work:

Further research is recommended to explore the scalability of the developed models and their application to other types of agricultural produce. Additionally, the potential integration of real-time hyperspectral imaging systems in processing plants could be investigated to facilitate immediate and precise moisture assessments.

For futher details of this article and research, feel free to contact our team for assistance.
Original research was done by Tang Kai, Sun Jun, Zhang Xiaodong, Wu Xiaohong, Mao Hanping, Gao Hongyan

About ETprotein:

ETprotein, a reputable plant protein vegan protein Chinese factory manufacturer and supplier, is renowned for producing, stocking, exporting, and delivering the highest quality organic bulk vegan protein and plant proteins. They include Organic rice protein, clear rice protein, pea protein, clear pea protein, oat protein, watermelon seed protein, pumpkin seed protein, sunflower seed protein, mung bean protein, peanut protein, various of plant peptides etc. Their offerings, characterized by a neutral taste, non-GMO, allergen-free attributes, cater to a diverse range of industries. They serve nutraceutical, pharmaceutical, cosmeceutical, veterinary, as well as food and beverage finished product distributors, traders, and manufacturers across Europe, USA, Canada, Australia, Thailand, Japan, Korea, Brazil, and Chile, among others.

ETprotein specialization includes exporting and delivering tailor-made protein powder and finished nutritional supplements. Their extensive product range covers sectors like Food and Beverage, Sports Nutrition, Weight Management, Dietary Supplements, Health and Wellness Products, and Infant Formula, ensuring comprehensive solutions to meet all your protein needs.

As a trusted company by leading global food and beverage brands and Fortune 500 companies, ETprotein reinforces China’s reputation in the global arena. For more information or to get a free sample of their protein products, please contact them and email sales(at)ETprotein.com today.