Early Identification and Localization Algorithm for Weak Seedlings Based on Phenotype Detection and Machine Learning

It is important to propose the correct decision for culling and replenishing seedlings in factory seedling nurseries improve quality of save resources. To solve problems inefficiency subjectivity existing traditional manual replenishment seeds, this paper proposes an automatic method discriminate early growth condition seedlings. Taking watermelon plug as example, Azure Kinect was used collect data its top view three times a day, at 9:00, 14:00, 19:00. The were collected from time germination main leaf growth, manually determined be strong or weak on last day collection. Pre-processing, image segmentation, point cloud processing methods performed obtain plant height area each seedling. sixth predicted using LSTM recurrent neural network first days. R squared prediction 0.932 0.901, respectively. dichotomous classification normal abnormal six machine learning methods, such random forest, SVM, XGBoost, data. experimental results proved that forest had highest accuracy 84%. Finally, appropriate decisions are given based results. This can provide some technical support theoretical basis transplanting robots.