Non-invasive Soya Bean Seed Analysis Using Machine Learning
Amrutha Patil1, Shashikumar G. Totad2
1Amrutha Patil, School of Computer Science and Engineering, KLE Technological University, Hubballi, Bangalore (Karnataka), India.
2Dr. Shashikumar G. Totad, School of Computer Science and Engineering, KLE Technological University, Hubballi, Bangalore (Karnataka), India.
Manuscript received on 06 February 2019 | Revised Manuscript received on 19 February 2019 | Manuscript Published on 04 March 2019 | PP: 279-282 | Volume-7 Issue-5S2 January 2019 | Retrieval Number: ES2047017519/19©BEIESP
Open Access | Editorial and Publishing Policies | Cite | Mendeley | Indexing and Abstracting
© The Authors. Blue Eyes Intelligence Engineering and Sciences Publication (BEIESP). This is an open access article under the CC-BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/)
Abstract: The soya bean is economically the most important legume in the world. Therefore, it is important to grow good quality seeds for a better yield. Identifying the right set of seeds is a difficult task when done manually since, there are no definite external characteristics of soya bean that correlate with its germination potential. Therefore, in this work an attempt is made at correlating the physical properties of soya bean with its germination potential using the concepts of machine learning and image processing. The input here being images, there are different methods to take images of soya bean, that is by using digital camera or radiography. The pros and cons of these methods are discussed. Since, using radiography images is not cost-efficient and its local availability for research purpose is scarce, a digital camera is used to take soya bean images. Once the image dataset is available, different classification methods are employed to classify the images into ‘germinating’ and ‘non-germinating’ seeds. The classifiers used are CNN, KNN and SVM and the average accuracy of the classifiers is 66.17%. The performance of different classifiers is analyzed to find the most suitable classifier. It is observed that most of the ‘germinating’ seeds have intact seed coat, elongated spherical shape, smooth texture and are evenly colored. Whereas, the other half has damaged seed coat, flat shape or not completely spherical, are unevenly textured and discolored at parts. Finally, the suggestions are made to improvise the results.
Keywords: CNN, Germination Potential, KNN, Machine Learning, Non-invasive, Radiography, Seed Analysis, Soya Bean, SVM.
Scope of the Article: Machine Learning