Integrating 3D detection networks and dynamic temporal phenotyping for wheat yield classification and prediction

Automated phenotyping of wheat growth stages from 3D point clouds is still limited. The study presents a concise framework that reconstructs multi-view UAS imagery into 3D point clouds (jointing to maturity) and performs plot-level phenotyping. A novel 3D wheat plot detection network—integrating spatial–channel coordinated attention and area attention modules—improves depth-direction feature recognition, and a point-cloud-density-based row segmentation algorithm enables planting-row-scale plot delineation. A supporting software system facilitates 3D visualization and automated extraction of phenotypic parameters. We introduce a dynamic phenotypic index of five temporal metrics (growth stage, slow growth stage, height/area reduction stage, maximum height/area difference stage, and height/area change rate) for growth-stage classification and yield prediction using static and time-series models. Experiments show strong agreement between predicted and measured plot heights (R 2 = 0.937); the detection net achieved AP 3D = 94.15 % and AP BEV = 95.35 % in “easy” mode; and a Bi-LSTM incorporating dynamic traits reached 82.37 % prediction accuracy for leaf area and yield, a 6.14 % improvement over static-trait models. This workflow supports high-throughput 3D phenotyping and reliable yield estimation for precision agriculture. • Developed a novel 3D wheat plot detection net with spatial–channel coordinated attention and area-attention modules, reaching 94.15% AP 3D and 95.35% AP BEV in high-precision mode, outperforming traditional methods. The CFPT 3D module boosts depth-direction feature extraction for dense planting. • Introduced 5 temporal phenotypic metrics (e.g., growth stage transitions, height/area change rates) to capture dynamic growth patterns. • Bi-LSTM models using these traits predicted yield with 82.37% accuracy, 6.14% higher than static-trait models. • Released a PyQt5-based 3D phenotype extraction tool for automated parameter calculation (height, canopy area, LAI) and visualization. • Proposed a density-based row segmentation algorithm enabling accurate row-level phenotyping, validated in single- and multi-row systems.