Measuring crop biomass is an integral process in agriculture, as it directly indicates the growth status of the crop. However, the traditional biomass measurement process is laborious, time-consuming, and requires manual sampling and weighing. To address this problem, we propose an innovative automated biomass prediction system using RGB-D cameras. We validate a range of biomass analysis techniques based on RGB-D data, encompassing conventional statistical models and machine learning approaches. These include the volume-based approach, support vector machine (SVM), random forest (RF), and extreme gradient boosting (XGBoost). Notably, we pioneer applying a deep learning-based 3D point cloud analysis model with RGB-D data for biomass prediction. We employ a 3D reconstruction algorithm to convert color and depth information into a 3D point cloud. Then, a biomass prediction model named Bio-DGCNN is developed based on Dynamic Graph CNN (DGCNN). Additionally, transfer learning is incorporated into the proposed system to improve the accuracy of biomass prediction further and make the system adaptable to tasks with small datasets. The experimental results demonstrate the effectiveness of our proposed system in accurately predicting biomass, outperforming other competitive approaches with small datasets. As a result, the proposed system provides a viable solution for farmers to make accurate biomass predictions even with limited datasets.