Development of an Efficient Grading Model for Maize Seedlings Based on Indicator Extraction in High-Latitude Cold Regions of Northeast China

Development of an Efficient Grading Model for Maize Seedlings Based on Indicator Extraction in High-Latitude Cold Regions of Northeast China

Maize, the world’s most widely cultivated food crop, is critical in global food security. Low temperatures significantly hinder maize seedling growth, development, and yield formation. Efficient and accurate assessment of maize seedling quality under cold stress is essential for selecting cold-toler...

Full description

Saved in:
Bibliographic Details
Main Authors: Song Yu, Yuxin Lu, Yutao Zhang, Xinran Liu, Yifei Zhang, Mukai Li, Haotian Du, Shan Su, Jiawang Liu, Shiqiang Yu, Jiao Yang, Yanjie Lv, Haiou Guan, Chunyu Zhang
Format: Article
Language:English
Published: MDPI AG 2025-01-01
Series:Agronomy
Subjects:
maize
low temperature stress
seedling quality
phenotypic evaluation indicators
efficient extraction
deep learning
Online Access:https://www.mdpi.com/2073-4395/15/2/254
Tags: Add Tag
No Tags, Be the first to tag this record!
Description
Summary:Maize, the world’s most widely cultivated food crop, is critical in global food security. Low temperatures significantly hinder maize seedling growth, development, and yield formation. Efficient and accurate assessment of maize seedling quality under cold stress is essential for selecting cold-tolerant varieties and guiding field management strategies. However, existing evaluation methods lack a multimodal approach, resulting in inefficiencies and inaccuracies. This study combines phenotypic extraction technologies with a convolutional neural network–long short-term memory (CNN–LSTM) deep learning model to develop an advanced grading system for maize seedling quality. Initially, 27 quality indices were measured from 3623 samples. The RAGA-PPC model identified seven critical indices: plant height (<i>x</i><sub>1</sub>), stem diameter (<i>x</i><sub>2</sub>), width of the third spreading leaf (<i>x</i><sub>11</sub>), total leaf area (<i>x</i><sub>12</sub>), root volume (<i>x</i><sub>17</sub>), shoot fresh weight (<i>x</i><sub>22</sub>), and root fresh weight (<i>x</i><sub>23</sub>). The CNN–LSTM model, leveraging CNNs for feature extraction and LSTM for temporal dependencies, achieved a grading accuracy of 97.57%, surpassing traditional CNN and LSTM models by 1.28% and 1.44%, respectively. This system identifies phenotypic markers for assessing maize seedling quality, aids in selecting cold-tolerant varieties, and offers data-driven support for optimising maize production. It provides a robust framework for evaluating seedling quality under low-temperature stress.
ISSN:2073-4395

Similar Items