Grading and Defect Detection in Potatoes Using Deep Learning
| Autor: | Nikhil Pandey, Suraj Kumar, Raksha Pandey |
|---|---|
| Rok vydání: | 2018 |
| Předmět: |
Deep cnn
Computer science business.industry Deep learning food and beverages Pattern recognition 02 engineering and technology 010501 environmental sciences Object (computer science) 01 natural sciences Transformation (function) 0202 electrical engineering electronic engineering information engineering Computer vision algorithms 020201 artificial intelligence & image processing Segmentation Artificial intelligence Grading (education) business Transfer of learning 0105 earth and related environmental sciences |
| Zdroj: | Communications in Computer and Information Science ISBN: 9789811323713 |
| DOI: | 10.1007/978-981-13-2372-0_29 |
| Popis: | Deep learning has been employed in a number of tasks. Taking inspiration from detection of tumors in medical tiff images, we had an idea of doing the same with other objects such as vegetables and plants which get affected by disease very often and still there are not many feasible approach of its detection. In this paper we present a practical approach to grade potato and classify the defects that might be present. We have used U-Net for segmentation of image containing on an average 50–60 potatoes. A physical object (marker) of known length is present along the potatoes for length reference. The U-net segmented result is then processed by computer vision algorithms (Distance transformation and watershed) to get the actual skin of the object of interest. After that we have used transfer learning to classify the skin for a number of defects such as greening, mechanical defect, rotting, sprouting etc. |
| Databáze: | OpenAIRE |
| Externí odkaz: |
|