DEVELOPING DEFECT INSPECTION MODEL OF HARICOT BEANS USING  COMPUTER VISION AND MACHINE LEARNING APPROACH

MELKAM, WALIE BALEW

dc.contributor.author	MELKAM, WALIE BALEW
dc.date.accessioned	2022-03-18T06:34:12Z
dc.date.available	2022-03-18T06:34:12Z
dc.date.issued	2021-10
dc.identifier.uri	http://ir.bdu.edu.et/handle/123456789/13210
dc.description.abstract	Haricot bean is a very important legume growing worldwide with higher market demands. It is a very important foreign exchange earning crop in Ethiopia. It has been observed that the markets of this crop are gradually increasing. In Ethiopia, the defect inspections are performed by experts manually. It is labour-intensive, time-consuming, and suffers from the problem of inconsistency and inaccuracy. In this study, we develop a model to inspect defect detection of haricot beans using computer vision and machine learning approaches. The required images of haricot beans were captured from Bure Ethiopian Commodity exchange (ECX) centre in the Amhara region of Ethiopia. 1000 for each class defect and non-defect haricot bean were taken. The images were taken directly using smart phone by placing on the white paper. After image acquisition pre-processing had been used to get an enhanced image. For feature extraction the grey level co-occurrence matrix (GLCM) and the convolutional neural network (CNN) method had been considered. Besides, for classification three classifiers random forest (RF) , support vector machine (SVM) and end to end CNN were applied to classify to their predefined class. For developing a prototype and conducting experiment, Python programming language was used in this study. In this study, three groups of experiment have been conducted (CNN features with SVM, RF; GLCM features with SVM, RF; finally and end to end CNN). From the experiment, the result revealed that the CNN method for feature extraction achieved an accuracy of 94% and 97% using SVM and RF classifiers, respectively. Further, using GLCM textural features methods were showed an accuracy of 88% and 97% for SVM, and RF classifiers, respectively. When using CNN as a classifier, an accuracy of 99% was achieved. It was concluded that, in all applied approaches, the model can identify defects and non-defect haricot bean with the highest accuracy. It is recommended that the developed approach should be implemented to other types of haricot beans, such as white beans and speckled beans. Key words: GLCM, CNN, SVM, RF, haricot bean.	en_US
dc.language.iso	en_US	en_US
dc.subject	INFORMATION TECHNOLOGY	en_US
dc.title	DEVELOPING DEFECT INSPECTION MODEL OF HARICOT BEANS USING COMPUTER VISION AND MACHINE LEARNING APPROACH	en_US
dc.type	Thesis	en_US