拉曼光谱结合智能算法在耐盐碱粳稻鉴定中的研究进展OA
Research Progress on the Identification of Salt-Alkali Tolerant Japonica Rice Using Raman Spectroscopy Combined with Intelligent Algorithms
盐碱化耕地制约粮食产能,寒区耐盐碱粳稻种质精准筛选是拓展耕地资源的核心环节.传统田间表型鉴定与分子标记辅助育种存在周期长、成本高、通量低等局限,难以满足规模化种质快速筛查需求.拉曼光谱具有无损、快速、分子指纹清晰、无需复杂前处理等优势,结合机器学习方法可实现粳稻耐盐碱性高通量智能判别,已成为作物抗逆种质鉴定的前沿技术.本文系统检索近 20 年发表的 59 篇中英文文献,明确检索策略与纳排标准,以拉曼光谱-智能算法耦合鉴定为主线,梳理三类粳稻耐盐碱性鉴定技术体系,阐述光谱采集、预处理、特征提取标准化流程;梳理了主流机器学习算法的适用场景、优势与局限并横向对比性能;总结了光谱成像融合不足、小样本泛化性差、田间便携检测配套缺失等瓶颈问题;展望多波段拉曼成像、一维光谱二维化、轻量化深度学习模型、田间原位智能检测等研究方向.综上,拉曼光谱-智能算法已具备作为耐盐碱粳稻种质前端高通量初筛工具的性能基础,可为寒地育种早期淘汰大量敏感材料提供快速无损的技术手段.
Saline-alkali cropland constrains grain productivity,and precise screening of saline-alkali-tolerant japonica rice germplasm in cold regions is a core step in expanding arable land resources.Traditional field phenotyping and molecular marker-assisted breed-ing have limitations such as long cycles,high costs,and low throughput,making them inadequate for large-scale rapid germplasm screening.Raman spectroscopy offers advantages including non-destructive and rapid detection,clear molecular fingerprinting,and no complex pretreatment requirement.Coupled with machine learning,it enables high-throughput intelligent discrimination of saline-al-kali tolerance in japonica rice and has become a frontier technology for crop stress-resistance germplasm identification.This paper systematically retrieved 59 publications in Chinese and English from the past 20 years,with clearly defined search strategies and in-clusion/exclusion criteria.Focusing on the coupling of Raman spectroscopy with intelligent algorithms for saline-alkali tolerance iden-tification in japonica rice,it reviewed three types of technical systems for identifying saline-alkali tolerance in japonica rice;described standardized workflows for spectral acquisition,preprocessing,and feature extraction;summarized the applicable scenarios,advan-tages,and limitations of mainstream machine learning recognition algorithms,and compared their performance;highlighted key bottle-necks such as insufficient spectroscopy-imaging fusion,poor small-sample generalization,and lack of field-portable detection sup-port;and envisioned directions including multiband Raman imaging,conversion of one-dimensional spectra into two-dimensional rep-resentations,lightweight deep learning models,and intelligent in situ field detection.Overall,Raman spectroscopy coupled with intel-ligent algorithms has demonstrated the foundational performance required for high-throughput preliminary screening of saline-alkali tolerant japonica rice germplasm,enabling rapid and non-destructive elimination of a large number of sensitive accessions at early stages of breeding in cold regions.
马波;王兴宇;胡继芳;杜浩强;刘蕊;李宸珲
黑龙江省农业科学院 齐齐哈尔分院,黑龙江 齐齐哈尔 161006黑龙江省农业科学院 齐齐哈尔分院,黑龙江 齐齐哈尔 161006黑龙江省农业科学院 齐齐哈尔分院,黑龙江 齐齐哈尔 161006黑龙江省农业科学院 齐齐哈尔分院,黑龙江 齐齐哈尔 161006齐齐哈尔大学,黑龙江 齐齐哈尔 161006齐齐哈尔大学,黑龙江 齐齐哈尔 161006
农业科技
耐盐碱粳稻拉曼光谱机器学习种质鉴定智能育种
saline-alkali tolerant japonica riceRaman spectroscopymachine learninggermplasm identificationintelligent breeding
《天津农业科学》 2026 (8)
55-64,10
农业科技创新跨越工程(CX23ZD02-03)黑龙江省重点研发计划(创新基地)(JD2023GJ04)黑龙江省优秀青年教师基础研究支持计划(YQJH2023065)黑龙江省自然科学基金(LH2024F046)
评论