Plant phenotyping methods are essential tools for the production of demanded crops for a growing population. Access to large-scale high-throughput screening systems are a critical barrier for early quantification of plants states in real-time. For this purpose, we developed an automated high throughput plant screening system which consists of two imaging chambers. The first chamber is equipped with two RGB cameras (top and side view), and a near-infrared hyperspectral imaging (NIR-HSI) system in the wavelength range of 900-1700 nm installed in the second screening chamber. Initially, the system was tested for early detection of chilling stress watermelon plants, a total of 350 watermelon plants were scanned before and after exposure to chilling stress conditions (-5℃) while moving through the chambers on the conveyor belt. An automatic image processing algorithm was developed for image segmentation and data augmentation. The color images applied to a transfer leaning of ResNet50 basis resulted in 90% classification accuracy. While the hyperspectral images were used to extract from each single plant leaves for the development of a partial least square discrimination analysis (PLS-DA) model which displayed over a 95% accuracy for the validation set. The overall results highlight that the high-throughput screen of plants on a combination of machine learning and deep learning has potential to quantify the plants' states under chilling stress condition.