LiDAR for measuring growth of plant in smart farm

Despite being of great importance to the economy and society of Thailand, the agricultural sector is unable to generate as much income as it should. The low average growth rate of agricultural production per labor is the main problem. Due to traditional farming practices, the resulting yields are only as effective as they should be, as it cannot compete with countries that are able to effectively apply digital technology and innovation. This study aims to address this issue by utilizing LiDAR (Light Detection and Ranging) technology and machine learning which used for predicting the accuracy of plant stage prediction to monitor the height and color of plants at three stages which divided into three stages, consisting of stage 1 for plants aged 0-7 days, stage 2 for plants aged 7-14 days, and stage 3 for plants aged 15-45 days to accurately determine their harvesting stage. This concept was demonstrated using LiDAR measurements of lettuces that consist of Green Cos, Green Oak, Butterhead, and Frillice Iceberg lettuces. Point clouds were generated from 3D RGB images and depth information from LiDAR camera (Intel RealSense L515) at three stages of the plant. For machine learning, feature extraction and model training and evaluation were used. Results show height feature prediction is 80% accuracy, RGB image feature prediction is 100% and height with RGB image feature prediction is 90%. This can be further applied to real-world use in smart farming in the future.