A study on the non-contact measurement of sunflower disk inclination and its application to accurate phenotypic analysis.

The tilt angle of sunflower flower heads is an important phenotypic characteristic that influences their growth and development, as well as the efficiency of mechanised harvesting in precision agriculture. Addressing the issues of low accuracy, high cost, and the risk of plant damage associated with traditional manual measurement methods, this study proposes a non-contact measurement method combining deep learning and geometric analysis to achieve precise measurement of sunflower flower head tilt angles. The specific method involves optimising the lightweight YOLO11-seg model to enhance instance segmentation performance for sunflower flower heads and stems (compared to the initial YOLO11 model, recall rate improved by 3.7%, mAP50 improved by 1.8%, a reduction of 0.29M parameters, and a decrease in computational load of 0.5 GFLOPs), and extracting the surface contour of the flower head and the centreline contour of the stem based on the mask map output by the model. After achieving precise region segmentation through image processing, the geometric analysis module performs elliptical fitting on the flower head contour to obtain the main axis direction, performs curve fitting on the stem contour, and selects the tangent direction at the intersection point of the flower head. The angle between the two is calculated as the tilt angle of the flower head. In the measurement experiment, 220 images were used for testing, with manual protractor measurement results as the reference. The algorithm achieved a measurement accuracy of RMSE = 2.93°, MAE = 2.43°, and R = 0.94. The results indicate that this method significantly improves measurement efficiency and operational convenience while maintaining accuracy. The system does not require contact with the plant, demonstrating good accuracy, adaptability, and practicality. The tilt angle information obtained is of great significance for path planning of harvesting robots, adjustment of gripping postures, and positioning control of end-effectors, and can serve as a key perception module in the automation process of sunflower flower head placement and drying operations in precision agriculture.