Automated detection and enumeration of planting mounds on images acquired by drone using deep learning

Reforestation of hybrid poplars on planting mounds is a technique recommended in Quebec to facilitate the rapid growth of the tree. However, in order to facilitate reforestation planning and operations, it is important to know the precise number of planting mounds when transporting new plants to the land to be reforested. Drone-acquired images were used to develop an automatic planting mound counting method. The method, based on computer vision and deep learning, can detect mounds with an average accuracy of 95.5 %, under various images acquisition conditions and on a wide variety of terrain types. The developed method has been successfully replicated in operational settings, demonstrating its robustness and optimizing reforestation planning efforts.