Comparison of Different Classifiers and the Majority Voting Rule for the Detection of Plum Fruits in Garden Conditions
| Autor: | Sajad Sabzi, Razieh Pourdarbani, Mario Hernández-Hernández, José Miguel Molina-Martínez, Davood Kalantari, Ginés García-Mateos, José Luis Hernández-Hernández |
|---|---|
| Přispěvatelé: | Universidad Politécnica de Cartagena, Universidad de Murcia, University of Mohaghegh Ardabili, Universidad Autónoma de Guerrero |
| Rok vydání: | 2019 |
| Předmět: |
Majority rule
Computer science Plum segmentation Producción Vegetal 3308 Ingeniería y Tecnología del Medio Ambiente Color space 01 natural sciences k-nearest neighbors algorithm remote sensing in agriculture Majority voting 5102.01 Agricultura Segmentation plum segmentation Environmental conditions lcsh:Science Tecnologías del Medio Ambiente Artificial neural network Pixel majority voting business.industry 010401 analytical chemistry Pattern recognition 04 agricultural and veterinary sciences Linear discriminant analysis artificial neural network hybridization Remote sensing in agriculture Artificial neural network hybridization 0104 chemical sciences environmental conditions Support vector machine Edafología y Química Agrícola 040103 agronomy & agriculture 0401 agriculture forestry and fisheries General Earth and Planetary Sciences lcsh:Q Artificial intelligence business |
| Zdroj: | Repositorio Digital de la Universidad Politécnica de Cartagena Fundación Universitaria San Pablo CEU (FUSPCEU) Remote Sensing; Volume 11; Issue 21; Pages: 2546 Remote Sensing, Vol 11, Iss 21, p 2546 (2019) |
| Popis: | Color segmentation is one of the most thoroughly studied problems in agricultural applications of remote image capture systems, since it is the key step in several different tasks, such as crop harvesting, site specific spraying, and targeted disease control under natural light. This paper studies and compares five methods to segment plum fruit images under ambient conditions at 12 different light intensities, and an ensemble method combining them. In these methods, several color features in different color spaces are first extracted for each pixel, and then the most effective features are selected using a hybrid approach of artificial neural networks and the cultural algorithm (ANN-CA). The features selected among the 38 defined channels were the b* channel of L*a*b*, and the color purity index, C*, from L*C*h. Next, fruit/background segmentation is performed using five classifiers: artificial neural network-imperialist competitive algorithm (ANN-ICA); hybrid artificial neural network-harmony search (ANN-HS); support vector machines (SVM); k nearest neighbors (kNN); and linear discriminant analysis (LDA). In the ensemble method, the final class for each pixel is determined using the majority voting method. The experiments showed that the correct classification rate for the majority voting method excluding LDA was 98.59%, outperforming the results of the constituent methods. This research was funded by the Spanish MICINN, as well as European Commission FEDER funds, under grant RTI2018-098156-B-C53. This project has also been supported by the European Union (EU) under Erasmus+ project entitled "Fostering Internationalization in Agricultural Engineering in Iran and Russia" [FARmER] with grant number 585596-EPP-1-2017-1-DE-EPPKA2-CBHE-JP. |
| Databáze: | OpenAIRE |
| Externí odkaz: |
|