A Novel Method of Identifying Paddy Seed Varieties.

Kuo-Yi Huang,Mao-Chien Chien

doi:10.3390/s17040809

Abstract

This paper presents a novel method for identifying three varieties (Taikong 9, Tainan 11, and Taikong 14) of foundation paddy seeds. Taikong 9, Tainan 11, and Taikong 14 paddy seeds are indistinguishable by inspectors during seed purity inspections. The proposed method uses image segmentation and a key point identification algorithm that can segment paddy seed images and extract seed features. A back propagation neural network was used to establish a classifier based on seven features that could classify the three paddy seed varieties. The classification accuracies of the resultant classifier for varieties Taikong 9, Tainan 11, and Taikong 14 were 92.68%, 97.35% and 96.57%, respectively. The experimental results indicated that the three paddy seeds can be differentiated efficiently using the developed system.

Highlights

Paddy is one of the main crops in Taiwan and can be planted twice each year
Purity analysis is crucial for nurseries and farmers, and purity is determined by paddy variety inspection
The results demonstrated that a backpropagation neural network (BPNN) with 20 hidden layer nodes, a learning rate

Summary

Introduction

Paddy is one of the main crops in Taiwan and can be planted twice each year. Purity analysis is crucial for nurseries and farmers, and purity is determined by paddy variety inspection. Purity is defined by professional inspectors according to the paddy’s appearance, shape, and color. There are approximately 500 cases (every case including approximately 4000 paddy seeds) of incorrect purity analyses every year in Taiwan. The jobs of inspection burden the inspectors with loading. Because healthy seedlings from seedling propagation stations (nurseries) are used to cultivate fields of paddy, seed quality is a critical factor when growing seedlings

Objectives

Methods

Results