Khumbaron Kiranbala Kabui1 and K.A. Athmaselvi1*
1 Research Scholar, Centre of Excellence for Grain Sciences, NIFTEM-Thanjavur-613005, India
1*Associate Processor and Head, Centre of Excellence for Grain Sciences, NIFTEM-Thanjavur-613005, India
Introduction
Food grains such as rice play a significant role in the human and animal food chains. Grain suppliers and food industries are focusing more on the quality of grain supplied from fields and grain mills with no tolerance for grain contamination as a result of growing customer awareness of high-quality and nutritious food. Rice quality is determined by a number of factors, including physical (hardness, moisture content, size, bulk density), compositional (starch content, milling yield, viability, oil, and protein) factors, safety (microbial infection, mycotoxins, and foreign material odor) (Jayas & Singh, 2012). The food industry still relies on manual inspection for quality assessment, which is time-consuming, expensive, and easily affected by physiological factors, leading to arbitrary and inconsistent evaluation results. Non-destructive techniques for assessing the quality of grains such as Computer vision/Machine vision can overcome such limitations. Machine vision systems have been employed extensively in the food sector as a quick, economical, reliable, and even precise inspection method for the purposes of assessing product quality.
Machine vision system
Machine vision is the process of using a computer to perceive an item and recognize its optical properties in order to interpret data. Based on the light that is reflected from an object that is illuminated by natural or artificial light, we perceive an object and its optical properties (Jha, 2010).
The main parts of a machine vision system include detector (camera) that functions similarly to the eyes, computer which resembles the human brain, frame grabber board, which transforms image data from the camera into digital form, interface (cable), electromagnetic radiation/light source facilitating capturing of image, and image acquisition and processing software. The major processes involved in a machine vision task are given below.
- image acquisition, which deals with factors like lighting, camera, and digitizer,
- image processing, which includes preprocessing, segmentation, and feature extraction, and
- image understanding, which includes picture recognition and interpretation.
Applications in Rice quality evaluation
Quality monitoring during processing
Computer vision systems could facilitate on-line grain monitoring while processing. Chen et al., (2020) conducted a real-time machine vision-based approach to monitor the rice combine harvester’s grain breakage rate. The image processing methods generated the colour spectra of the kernel visuals in HSL space to distinguish between whole and broken grains and to determine the extent of breakage. Jahari et al., (2015) uses a machine vision system utilizing double lighting to assess the quality of harvested paddy during harvesting. The captured images were analysed and the unwanted materials and damaged grain present in the paddy during harvest were determined.
Grading
A machine vision system was utilized in conjunction with multiple metaheuristic classification algorithms to qualitatively assess milled rice grains (Zareiforoush et al., 2016). Based on colour images, four metaheuristic classification methods including artificial neural networks, decision trees, Bayesian Networks, and support vector machines were employed to categorize milled rice into four qualitative levels as high-processed sound grains, low-processed sound grains, high-processed broken grains, and low-processed broken grains. Kaur & Singh, (2013) identified the short, long, thin, round, and bold classes of rice using an SVM classifier. The classification criteria used included changes in the dimension such as length, width, and length/width ratio of single grain. The approach can also distinguish head grain rice from broken and brewer rice based on comparative differences in grain diameters.
Variety based classification
Manual classification of variations is increasingly a concern due to the availability of various new types. Using machine vision systems, researchers have developed algorithms that account for sample color and seed size information to explain grain varietal differences. Guzman & Peralta, (2008) developed multilayer neural networks and machine vision systems to automatically identify the five different varieties of rice grown in the Philippines (saline prone, lowland rain-fed, upland rice varieties, lowland irrigated, and cool elevated varieties) based on the sizes, shapes. The grain sizes and shapes could be identified by the Artificial Neural Network classifiers with mean accuracy of 96.67% and 98.76% respectively.
Identification of insect infestation
Identification and classification of insects is crucial to prevent infestation of stored grains. Mendoza et al., (2023) applied machine learning for detecting insects in grain facilities and was effective in accurately identifying and classifying warehouse beetles and adult cigarette beetles. Rathore et al., (2019) developed classification systems for the identification of granaries insects with the use of feature extraction and machine learning and it could identify the types of hidden adult and larval insects. Based on the attributes obtained, it distinguishes the sounds of insects that were recorded in a grain silo.
Detection of fungal infection
A major problem for the grain industry is microbial contamination. Food grains do not support bacterial development under typical storage conditions, however unfavorable temperature ranges and moisture content can encourage the growth of fungus species. Arinichev et al., (2021) detected rice fungal diseases using modern computer vision methods and reported that the disease can be identified with at least 95% accuracy in the dataset utilized for the purpose of the study.
Limitations
Machine vision systems have been primarily utilized to classify grains and their varieties based on physical appearance when assessing the quality of food grains. However, most machine vision systems lack the ability to understand the grain’s composition, organoleptic properties, and internal insect infestation, that are crucial determinants of food grain quality. Another challenge is the decision on the involved learning technique. Machines that use image acquisition and processing techniques need specialized manpower and have issues with time-consuming troubleshooting procedures. Even though computer-oriented methods can be designed, examined, and solved quickly, their application in real-time systems is frequently constrained due to greater computation needs (Vithu & Moses, 2016).
Conclusion
Machine vision systems offer safe, non-contact, and non-invasive ways to assess the quality of food grains. The technique may quickly and accurately offer information about the exterior quality characteristics of food grains. They can be employed for grain discoloration, insect infestation, fungal infection, and type and variety classification. The main difficulty is integrating such systems with those that can describe internal feature of grain quality. With recent developments in optical instruments, computer hardware, optical sensors, and fast processing methodologies, computer vision techniques could play a major role in grain quality and safety monitoring systems. In future, computer vision systems will be able to successfully address a variety of grain quality evaluation applications due to increasing application needs and research advancements.
References
Arinichev, I. V., Polyanskikh, S. V., Volkova, G. V., & Arinicheva, I. V. (2021). Rice Fungal Diseases Recognition Using Modern Computer Vision Techniques. International Journal of Fuzzy Logic and Intelligent Systems, 21(1), 1–11.
Chen, J., Lian, Y., Zou, R., Zhang, S., Ning, X., & Han, M. (2020). Real-time grain breakage sensing for rice combine harvesters using machine vision technology. International Journal of Agricultural and Biological Engineering, 13(3), 194–199.
Guzman, J. D., & Peralta, E. K. (2008). Classification of philippine rice grains using machine vision and artificial neural networks,. World Conference on Agricultural Information and IT, IAALD AFITA WCCA, November, 41–48.
Jahari, M., Yamamoto, K., Miyamoto, M., Kondo, N., Ogawa, Y., Suzuki, T., Habaragamuwa, H., & Ahmad, U. (2015). Double lighting machine vision system to monitor harvested paddy grain quality during head-feeding combine harvester operation. Machines, 3(4), 352–363.
Jayas, D. S., & Singh, C. B. (2012). Grain quality evaluation by computer vision. In Computer Vision Technology in the Food and Beverage Industries. Woodhead Publishing Limited.
Jha, S. N. (2010). Nondestructive evaluation of food quality: Theory and practice. In Nondestructive Evaluation of Food Quality: Theory and Practice.
Jinorose, M., Prachayawarakorn, S., & Soponronnarit, S. (2014). A novel image-analysis based approach to evaluate some physicochemical and cooking properties of rice kernels. Journal of Food Engineering, 124, 184–190.
Kaur, H., & Singh, B. (2013). Classification and Grading Rice Using Multi-Class SVM. International Journal of Scientific and Research Publications, 3(4), 624–628.
Mendoza, Q. A., Pordesimo, L., Neilsen, M., Armstrong, P., Campbell, J., & Mendoza, P. T. (2023). Application of Machine Learning for Insect Monitoring in Grain Facilities. Ai, 4(1), 348–360.
Rathore, D. S., Ram, B., Pal, B. L., & Malviya, S. (2019). Analysis of Classification Algorithms for Insect Detection using MATLAB. SSRN Electronic Journal.
Vithu, P., & Moses, J. A. (2016). Machine vision system for food grain quality evaluation: A review. Trends in Food Science and Technology, 56, 13–20.
Zareiforoush, H., Minaei, S., Alizadeh, M. R., & Banakar, A. (2016). Qualitative classification of milled rice grains using computer vision and metaheuristic techniques. Journal of Food Science and Technology, 53(1), 118–131.
