Anomaly detection in plant growth in a controlled environment using 3D scanning techniques and deep learning

This paper presents a comparison of different methodologies for monitoring the plants growth in a greenhouse. A 2D measurement based on Computer Vision algorithms and 3D shape measurements techniques (Structured light, LIDAR and photogrammetry) are compared. From the joined 2D and 3D data, an analysis was performed considering health plant indicators. The methodologies are compared among each other. The acquired data are then fed into Deep Learning algorithms in order to detect anomalies in plant growth. The final aim is to give an assessment on the image acquisition methodologies, selecting the most suitable to be used to create the Deep Learning model inputs saving time and resources.