Liangchao Deng (邓良超)

3D plant reconstruction, point cloud structural analysis, robotic sensing for high-throughput phenotyping, autonomous data collection

Multimodal AI applications in image segmentation, target detection, plant phenotypic analysis, and structure-function modeling

Physics-based canopy modeling and simulation for high-efficiency crop research

RGB, multispectral, hyperspectral imaging, and LiDAR data fusion for smart agriculture

Coupling sensing data with crop growth models to build agricultural automation simulation systems

Supervisors: Prof. Yali Zhang (Shihezi University); Dr. Qingfeng Song (CEMPS, CAS); Prof. Xin-Guang Zhu (CEMPS, CAS)

Research Focus: Crop phenomics, UAV remote sensing, canopy photosynthesis modeling, AI-assisted phenotyping

Core Projects:

- 3D crop canopy reconstruction and light distribution simulation: Based on multi-view stereo (SfM) and 3D Gaussian Splatting (3DGS) algorithms

- Multi-source sensor data fusion: Collaborative processing and analysis of RGB, multispectral, hyperspectral, and LiDAR data

- Machine learning in plant phenotypic extraction: Deep learning for image segmentation, target detection, and phenotypic parameter prediction

Joint Training: CAS Center for Excellence in Molecular Plant Sciences (CEMPS), participated in national-level research projects

Professional Foundation: Solid numerical analysis, computational modeling, programming, and algorithm design skills

Core Courses: Computer Vision, Machine Learning, Linear Algebra, Optimization Algorithms, Graph Theory, Data Structures and Algorithms

Graduation Project: Numerical simulation based on computational fluid dynamics (Excellent graduation project)

3D Crop Canopy Reconstruction: Implemented high-throughput 3D reconstruction of farmland scenes using SfM (Structure from Motion) and 3DGS (3D Gaussian Splatting) algorithms with UAV cross-circular data acquisition methods, achieving centimeter-level reconstruction accuracy.

Canopy Light Distribution and Photosynthesis Model: Established canopy-scale light distribution and photosynthesis models, optimized ray tracing and BRDF-based leaf optical properties, forming a digital twin framework for crop growth, with significantly improved canopy light distribution simulation accuracy compared to traditional models.

Multimodal AI Data Processing: Leveraged multimodal AI (RGB, multispectral, LiDAR) data processing capabilities to achieve complex-scene, zero-shot plant segmentation, providing a data analysis foundation for field-walking high-throughput phenotyping platforms.

Modular AI Agent Development: Transformed the canopy photosynthesis model based on AI, modularizing 3D reconstruction, meshing, canopyization, light distribution simulation, and photosynthesis calculation to build a crop canopy photosynthesis AI agent.

BRDF-based Leaf Optical Property Inversion Framework: Developed a leaf optical property inversion algorithm based on BRDF theory, optimized measurement schemes, and achieved indirect measurement of leaf optical properties.

High-Efficiency-Oriented Wheat Design Research: Conducted collaborative modeling studies on optimal plant architecture and cultivation configurations based on wheat canopy photosynthesis models, including industry-oriented projects (e.g., BASF), translating research into practical agricultural solutions to improve weed suppression efficiency.

Computer Vision Algorithm Development: Built multiple image processing and quantification methods based on computer vision algorithms for automatic extraction and analysis of plant phenotypic parameters.