Quality Testing of Food Grains Using Image Processing and Neural Network

Abstract

The main crop for our nation to boost agricultural income is grains. While the crop is still in the ground, farmers notice yield the most, but once the grain has been processed and sold, quality becomes the main factor in determining its viability. Several contaminants, including stones, weed seeds, chaff, damaged seeds, etc., are present in these grains. Low automation levels and a large human workforce are required for assessing grain quality. Additionally, it increases the cost and length of the testing process. This contradiction is becoming more and more obvious as import and export trade expands. Prior to performing the next process during grain handling procedures, several types of grain and their quality are necessary. For the identification of grain seed varieties and quality nowadays, we use scientific approaches. The scientific approach is also extremely labor-intensive and ruins the used sample. In contrast to the chemical approach, machine vision or digital image processing is a non-destructive method that is also a highly convenient and inexpensive process. In order to identify different types of grains and determine the purity of grains using image processing techniques based on various parameters including grain size and shape, we proposed a grain classification system based on machine learning and image processing algorithms. The Matlab programming language and Matlab software are used for all operations. Images are collected from a dataset that includes images of food grains. On the captured images, feature extractions, segmentation, and image processing techniques are applied. That can be extracted in a non-contact method from the grains. This paper will also discuss and offers suggestions for how to categorize different types of food grains. It also determines the purity of the grain using image processing techniques based on characteristics like major axis length, minor axis length, area, and others.

Introduction

I. INTRODUCTION

Food Grains is the most important food consumed by a huge population in Asian countries. Rice is a member of the Poaceae family of plants. Rice is grown in a variety of locations around the world. India is the world’s second-largest producer of rice. The demand for high-quality food grains rises in demand with the rise of its consumption. Food grains are differentiated using manual classification methods based on local geometric features in local industries.

The suggested work employs a method for processing recorded digital photographs of food grains and extracting significant information. Morphological characteristics are examined to establish the  type of food grain. To extract various information from the collected image, image processing techniques are used.

The food granules are evaluated by a neural network after picture processing. The results are acquired by putting the rice grains through a series of tests.

Using image processing and neural network technologies, this paper[5] performs quality grade testing and identification of rice granules. Images for rice are captured here using a webcam. MATLAB is used to conduct image pre-processing techniques such as Thresholding, segmentatione extraction on the obtained image. For training purposes, the features are supplied to the neural network. The trained network is then utilized to determine the quality of the unknown contaminants. The  grading system was created to ensure product quality consistency.

The Neural Network Pattern Recognition Tool may be used to classify granules using this way.  MATLAB provides many functions in the form of a toolbox that helps us in automating commonly used image processing techniques and workflows by enabling interactive segmentation of image data, comparison of image registration methods, and batch processing of large datasets. There are few toolbox that we have used in our project for front end part we have used GUIDE toolbox, for image processing we have utilised image processing toolbox and for training and testing we have used neural network toolbox.

II. LITERATURE SURVEY

A. Quality Testing of Food Grains

Authors: Sheikh Bilal Ahmed, Syed Farooz Ali, and Aadil Zia Khan, "On the Frontiers of Rice Grain Analysis, Classification and Quality Grading”.

Description:- This work organizes and categorizes rice grain classification and quality grading procedures into geometrical, statistical, supervised, unsupervised, and deep learning approaches in chronological sequence. They conclude in this paper that deep learning produces the greatest results. From 1996 to the present, this research examines the evolution of rice grain classification algorithms.

Authors: Nikhade Pratibha, Hemlata, "Analysis and Identification of Rice Granules Using Image processing and neural network”.                                 

Description:- This presented a study project based on Image processing and neural networks were used to establish the notion of quality grade testing and identification On the captured image, we use the image processing method in MATLAB to execute Image Pre-processing techniques, Ostu's Thresholding, Canny edge detection, and Feature extraction. A neural network is given with features as part of the training process. Unrecognized contaminants and quality are determined using the trained network. A PC webcam camera is used for the acquisition, which is done in uniform lighting. Smoothing collected images is done with a median filter. In our study, we use Otsu's thresholding technique to conduct thresholding. The quality of rice seeds is assessed using image processing. The size of the grains in the sample determines the grading of rice granules. Grade 1 describes approximately 55% of long grains. In the same way, 33.33 percent of tiny grains are classified as grade 2.

Authors: P. Neelamegam and S. Sudha, "Machine vision-based quality analysis of rice grains".

Description:- The web camera is used in this paper to inspect rice grains using image processing techniques. They employ the OTSU approach, Grayscale method, Bounding Box method, and Blob detection method in this paper. This paper employs a camera to inspect rice grain quality under two different conditions: non-overlapping and overlapping.

Experiments have shown that the system can function accurately and reliably in non-overlapping arrangement situations, with an average error of 0.47 percent. The mistakes from the overlapping arrangement tests were highly high in the testing under all conditions of 18 experimental designs, averaging 53.82 percent on average. However, as compared to the overlapping approach, the non-overlapping method produces less inaccuracy.

Authors: P.M Devi Analysis and Identification of Rice Granules Using Image Processing and Neural Network”.

Description:- Using image processing and neural network technologies, this paper performs quality grade testing and identification of rice granules. Images for rice are captured here using a webcam. MATLAB is used to conduct image pre-processing techniques such as Ostu's Thresholding, Canny edge detection, and feature extraction on the obtained image. For training purposes, the features are supplied to the neural network. The trained network is then utilized to determine the quality of the unknown contaminants. The grading system was created to ensure product quality consistency. The Neural Network Pattern Recognition Tool may be used to classify granules using this way.

Authors: L.Shobha, “Rice Quality Evaluation Based on Image Processing A Survey”.

Description:- A survey of image processing approaches utilized in automated rice grading systems in an agricultural context was described in this paper [8]. Image processing techniques such as background subtraction, feature extraction, training, and classification are used in this field. From the published literature, an image processing-based method for automatic rice detection, classification, and recognition of foreign particles from images of rice grains utilizing color and texture data is investigated. Image processing techniques have been widely used in agricultural applications. It has the potential to be a useful tool for determining the quality of food. For the realization of real-time requirements, there are numerous options.

Authors:  Oliver C. Agustin and Byung-Joo Oh, "Weight Estimation and Classification of milled rice using Support Vector Machine”.

Description:- For weight estimation and classification of milled rice kernels, this paper [2] use a Support Vector Machine. They developed a support vector regression (SVR) model for predicting rice kernel weight as well as a support vector classifier (SVC) for detecting rice defects. SVR outperformed linear regression (LR), according to the findings. The performance of the suggested SVR model was found to be superior to that of the LR model.

III. ARCHITECTURE AND DESIGN

A. System Overview

V. TESTING

The Testing was done using the neural network, We utilized MATLAB software for Training and testing. The code was written in matlab programming language.  MATLAB provides many functions in the form of a toolbox that helps us in automating commonly used image processing techniques and workflows by enabling interactive segmentation of image data, comparison of image registration methods, and batch processing of large datasets. There are few toolbox that we have used in our project:

1. Guide Toolbox: This toolbox offers resources for designing complex, gracefully resizable MATLAB graphical user interfaces. The provided classes can be combined to create practically any user interface design.
2. Image Processing Toolbox: Perform image processing, visualization and analysis on the images in the dataset. You can perform image segmentation, image enhancement, noise reduction, geometric transformations, and image registration using deep learning and traditional image processing techniques. The toolbox supports processing of 2D, 3D, and arbitrarily large images.
3. Neural Network Toolbox: A neural network (also called an artificial neural network) is an adaptive system that learns by using interconnected nodes or neurons in a layered structure that resembles a human brain. A neural network can learn from data—so it can be trained to recognize patterns, classify data, and forecast future events. This toolbox is used to train images and get the results.

We  tested on the around 200 images, each image containing different number of grains. During  the training, neural network weights are  initiated with random values after which  they were adjusted based on error back propagation  learning. The weights  are  stored during  the end of  training and  the stored weights are used at  the  time of testing. When  the  training has completed,  the network has been tested on 200 samples of food grain images.

VI. EXPERIMENTATION AND RESULTS

The  food granules  are  graded  by  Neural  Network  system  which  is designed  for  this purpose. The neural network  is  trained with area, major  axis, minor  axis  and  perimeter  of  granules.  The trained neural network is used to compute the grade by giving the grain input variables for the unknowable output (grades). The expectations of the expert are flexibly reflected by neural networks. Due to time constraints, the number of images captured for training and testing has been limited. For assessing the categorization of granule grade classification, 87 samples are employed. It can accurately classify all of the grains depicted  when there is no grain overlap. The accuracy in this scenario is found to be 96 percent since the Neural Network cannot accurately distinguish if there is grain overlap.

A. Features Extraction (TRAINING PHASE)

Enter Grain Type: 1. Rice 2. Wheat 3.Corn 4.Horse Gram 5. Impurity

Enter Grain Grade: 1. Basamati 2.Sona Masuri  

Select the Grain image for training phase

???????

Conclusion

The complexity of the grading problem was significantly decreased through image processing and careful selection of the species that were taken into consideration in this work for extracting features from rice grains. Grading rice particles using a neural network is successful. The created neural network can also be used to grade different grains and food items.When there is no granule overlap, the Probalistic based Neural Network can classify well, but when there is granule overlap, it can categorise the test datasets with 90 percent accuracy. We worked on the area detection on the rice grain and created an image processing system to grade the rice based on length, width, area, and area of chalky. Based on the findings, it can be said that some rice are better based on length, some are better based on breadth, and some may be considered to be of good quality based on area and area of the chalky. All of the traits need not, however, be represented in the rice grain. For further verification of our methods, additional data can be collected. The amount of moisture in a rice grain can be added to a grade to indicate the overall quality of the rice for more research. A. Future Works 1) The preliminary work presented in the paper could be further enhanced by focusing on different sampling methods, sample preprocessing techniques, different features, and different neural network model to match the requirements of the rice industry. 2) The quality of food grain includes not only exterior quality but also the interior one such as the texture, the nutrient component, the protein, moisture of the grain and producing area. The latter is not researched in this work because of the limitation of machine vision. It is difficult to test the features by visual image. 3) Even though the problem being worked upon is not completely new, the earlier approaches employed very large number of color, textural and morphological features which made the algorithm extremely slow because of the intensive computation. 4) An efficient method is proposed for classification of food grains which require limited features and thus overcoming the disadvantages like tediousness and time consumption.

References

[1] Sheikh Bilal Ahmed, Syed Farooz Ali, and Aadil Zia Khan, \"On the Frontiers of Rice Grain Analysis, Classification and Quality Grading. [2] Oliver C. Agustin and Byung-Joo Oh, \"Weight Estimation and Classification of milled rice using Support Vector Machine”. [3] Priyadarshini Patil, \"Reliable Quality Analysis of Indian Basmati Rice Using Image Processing”. [4] Archana Chaugule and Suresh N. Mali, Hindawi Publishing Corporation Journal of Engineering Volume 2014, Article ID 617263, 8 pages, “Evaluation of Texture and Shape Features for Classification of Four Paddy Varieties”. [5] International Journal of Electronics and Communication Engineering. a. ISSN 0974-2166 Volume 10, Number 1 (2017), pp. 25-33 © International Research Publication House,” Analysis and Identification of Rice Granules Using Image Processing and Neural Network”. [6] T. G. Devi, P. Neelamegam and S. Sudha, \"Machine vision-based quality analysis of rice grains,\" 2017 IEEE International Conference on Power, Control, Signals and Instrumentation Engineering (ICPCSI), 2017, pp. 1052-1055, DOI: 10.1109/ICPCSI.2017.8391871. [7] W. Sriratana, N. Narknam, R. Apichitanon, and N. Tammarugwattana, \"Application of Webcam for Inspection of Rice Grain Quality by Using Image Processing Technique,\" 2020 20th International Conference on Control, Automation and Systems (ICCAS), 2020, pp. 1134-1139, DOI: 10.23919/ICCAS50221.2020.9268316. [8] https://www.researchgate.net/publication/334941834 “Rice Quality Evaluation Based on Image Processing A Survey”. [9] Kittinun Aukkapinyo, Suchakree Sawangwong, Parintorn Pooyoi, and Worapan: International Journal of Automation and Computing, vol. 17, no. 2, pp. 233-246, 2020 DOI: 10.1007/s11633-019-1207-6, “Localization and Classification of Rice-grain Images Using Region Proposals-based Convolutional Neural Network”. [10] Anis Sufiya Hamzah, Azlinah Mohamed, IAES International Journal of Artificial Intelligence (IJ-AI) Vol. 9, No. 4, December 2020, pp. 600~608 ISSN: 2252-8938, DOI: 10.11591/ijai.v9.i4.pp600-608,” Classification of white rice grain quality using ANN”.

Copyright

Copyright © 2022 Nayana KB, Geetha CK. This is an open access article distributed under the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.