Plant phenotyping enables quantitative monitoring of crop performance and stress responses. The LI-COR LI-600 is the gold standard for non-destructive leaf-level physiological measurement. However, it is slow, labour-intensive, and costly, limiting its suitability for large-scale or high-frequency studies. Remote sensing via satellites or UAVs enables large-scale monitoring but faces resolution and environmental limitations. Proximal imaging, conducted within 1-2 metres of the plant canopy, allows substantially higher spatial and temporal resolution. This enables yield and biomass prediction and pest, disease, and weed identification, all are difficult to perform reliably at UAV or satellite scales. This thesis describes the design and evaluation of robotic High-Throughput Phenotyping Systems (HTPS) using proximal RGB and multispectral imaging. Two gantry-based robots were developed: a 3 × 6 m RGB HTPS for greenhouse imaging of potted oats, and an outdoor raised-bed HTPS using a custom multispectral camera to image oat and barley rows. Both systems were paired with LI-COR measurements; yield and biomass were measured outdoors. The image processing compared conventional plant phenotyping metrics with three deep learning models (ResNet50, Swin, EfficientNet-B5) to predict plant parameters. This thesis answers two key research questions: (1) Can proximal RGB imaging predict LI-COR physiological measurements for individual oat plants in a greenhouse setting? The SWIN transformer achieved a prediction R2 = 0.27 for leaf vapour pressure and R2 = 0.40 for leaf vapour pressure deficit, and an R2 = 0.41 for stomatal conductance. (2) Can proximal multispectral imaging estimate LI-COR physiological metrics, yield and biomass for oats and barley in outdoor field conditions? In field conditions, the SWIN transformer achieved a prediction R2 = 0.78 for leaf vapour pressure, an R2 = 0.67 for leaf vapour pressure deficit, and an R2 = 0.29 for stomatal conductance. Resnet predicted yield in oats with R2 = 0.52 and biomass in oats and barley combined with R2 = 0.56. Two new datasets were produced and will be publicly released, including the first documenting visible predation in cereal crops.