Summary
International Technical Conference on Circuits/Systems, Computers and Communications
2016
Session Number:W1-1
Session:
Number:W1-1-1
Machine Learning for Classifying Working State Images Recorded by Digital Tachograph System
Senlin Guan, Takeshi Shikanai, Morikazu Nakamura ,
pp.689-692
Publication Date:2016/7/10
Online ISSN:2188-5079
Summary:
Machine learning is a powerful modelling and prediction tool for data analysis and decision-making in agriculture production, especially for the cases of dealing with large volume of data in diverse formats. In this paper, we present a case study of applying machine learning to classify the working states of harvesting sugarcane based on the time-series data, which is recorded by a digital tachograph system mounted on a small sugarcane harvester. The study aims at constructing a model and training it applicable to automatically learn from the time-series images and classify the images into different working states. Three machine learning models are implemented to evaluate the best accuracy for classification and the optimum parameters for the model. The result indicates that using machine learning is an effective way to distinguish the working states, and the average F1_score reaches 0.970 when recognizing the cutting state. The classification by the Support Vector Machine (SVM) model with Radial Basis Function (RBF) kernel gains higher accuracy than by that with linear kernel and by K-Nearest Neighbors (KNN).