This work presents and compares several methods for automatic plant Height Estimation in a controlled agricultural environment. The proposed approaches involve the detection and bounding of the plant using either image processing based on color segmentation or deep learning method based on the YOLO family architectures. Once the plant is framed, the pixel-to-centimeter conversion is performed using either a geometric model or a corrected formulation that accounts for camera distortions and tilt angle. Experiments were conducted with both horizontal and inclined camera setups, varying the distance and height of the device relative to the plant. The best results were achieved using the neural network combined with the direct conversion method, particularly when the camera was positioned at 3.5 times the plant height and inclined at approximately 25 degrees. This configuration yielded a Mean Absolute Percentage Error as low as 1.61% and a Mean Absolute Error of 1.01 cm. The proposed system shows potential for replacing manual measurements in precision agriculture, enabling real-time monitoring of plant development and optimizing resource management.

DOI: https://doi.org/10.1109/MetroAgriFor66923.2025.11512364