Research & Projects 科研与项目经历

Research & Projects

Multi-modal Intelligent Monitoring of the Full Wheat Phenology

小麦全物候期多模态智能监测

2025.05 – 2026.05
  • Algorithm design: Built an image visual-feature extractor on a Transformer architecture, modelled environmental information with sequence models such as LSTM, fused multi-modal features through MLPs, and designed a plot segmentation algorithm to achieve accurate, intelligent recognition of wheat phenological stages.
  • Core work: Constructed a full-phenology monitoring dataset for wheat from UAV RGB imagery and contributed to a multi-modal data-fusion algorithm that integrates environmental information.
  • Outcome: The project passed its national-level completion review, delivering the wheat phenology dataset and the multi-modal fusion algorithm.
  • 算法搭建:基于 Transformer 架构构建图像视觉特征提取器,使用 LSTM 等序列模型建模环境信息,基于 MLP 等融合多模态特征,并设计小区分割算法,实现小麦物候期的精准智能识别。
  • 核心工作:基于无人机 RGB 图像构建小麦全物候期监测数据集,参与开发融合环境信息的多模态数据融合算法。
  • 项目成果:项目以国家级结项验收,完成小麦全物候期监测数据集构建与多模态数据融合算法开发。
  • Transformer
  • LSTM
  • MLP FusionMLP 融合
  • UAV RGB无人机 RGB
  • Phenology Recognition物候期识别

GWAS of Yield-Related Traits in Rice and Identification of Small-Effect Polygenes

水稻产量性状 GWAS 与微效多基因识别

2025.09 – 2026.02
  • Algorithm design: Using compressed variance-component mixed models, brought additive and dominance effects together with their polygenic background into a unified Bayesian framework, and applied a two-stage strategy of genome-wide scanning plus machine learning to identify small-effect polygenes accurately from sparse solutions.
  • Core work: Applied statistical genetic models to genome-wide association studies (GWAS), functional enrichment and candidate gene discovery for yield-related traits in rice.
  • Outcome: Enabled the mining of elite germplasm carrying pleiotropic genes in rice; the results were applied in a 2026 article in Plants (JCR Q1, IF 4.7).
  • 算法搭建:通过压缩方差组分混合模型,将加性、显性效应及其多基因背景纳入统一的 Bayesian 框架,并利用全基因组扫描 + 机器学习的两阶段策略,从稀疏解中精准识别微效多基因。
  • 核心工作:使用统计遗传模型对水稻产量性状进行全基因组关联分析(GWAS)、功能富集与候选基因寻找。
  • 项目成果:完成水稻一因多效基因的优良种质挖掘,研究成果已应用于 Plants(JCR Q1, IF 4.7)2026 年论文。
  • Bayesian FrameworkBayesian 框架
  • GWAS
  • Variance-Component Model方差组分模型
  • Candidate Genes候选基因

PhenoNet — An Integrated Air–Ground Wheat Phenology Platform

PhenoNet — 空地一体化小麦物候解析平台

2025.08 · Kunming, Yunnan 2025.08 · 云南昆明
  • Overview: Took part in discussions on the frontiers of plant phenomics, focusing on three pain points — the scarcity of wheat phenology datasets, the low accuracy of manual collection and the steep learning curve of analysis algorithms.
  • Outcome: Presented on behalf of team AiPhenomics, winning the National Third Prize in the undergraduate track.
  • 项目概述:参与交流植物表型组学前沿进展,聚焦小麦物候数据集匮乏、人工采集精度低、分析算法上手难三大痛点。
  • 项目成果:代表 AiPhenomics 团队作报告,获本科生赛道国家三等奖。
  • Smart Agriculture智慧农业
  • Phenomics表型组学
  • Platform Development平台建设

Real-time Emotion Detection System: A Lightweight CNN Approach

2026.02 · Singapore 2026.02 · 新加坡
  • Overview: Built a real-time emotion detection system on a CNN framework, using Conv2D for feature extraction and MaxPooling2D for feature compression to deliver emotion detection under a lightweight budget.
  • Outcome: Led the project on behalf of the team and was graded DISTINCTION individually.
  • 项目概述:基于 CNN 框架构建实时情绪检测系统,利用 Conv2D 提取特征、MaxPooling2D 压缩特征,在轻量化前提下实现情绪检测功能。
  • 项目成果:代表团队主持该项目,个人获评 DISTINCTION。
  • CNN
  • Real-time Inference实时推理
  • Lightweight Model轻量化模型

Mathematical Modelling: Trajectory Computation, Collision Avoidance and Speed Optimisation

数学建模:轨迹计算、碰撞避免与速度优化

2024.11
  • Contribution: Helped build equidistant-spiral and single-objective optimisation models, solving trajectory computation, collision avoidance, turnaround path design and speed optimisation for the motion of a dragon-dance team.
  • 核心工作:参与构建等距螺线、单目标优化等模型,解决舞龙队运动中的轨迹计算、碰撞避免、调头路径设计及速度优化等问题。
  • Equidistant Spiral等距螺线
  • Single-objective Optimisation单目标优化
  • Operations Research运筹学