Corp breeding plays an important role in modern agriculture, improving plant adaptability and increase yield. Optimizing genes is the key step to discover the beneficial genetic traits for crop production increasing. Associating genes and their functions needs a mountain of observation and measurement of the phenotypes, which is a dreary and fallible job for human beings. Automatic seedling phenotyping system aims at replacing the manual measurement, reduce the sampling time and increase the allowable work time. In this research, we developed an automated maize seedling phenotyping platform based on a ToF camera and an industrial robot arm. A ToF camera is mounted on the end-effector of the robot arm. The arm brings ToF camera to different viewpoints for acquiring 3D data. Camera-to-arm transformation matrix is calculated from hand-eye calibration, which is applied to transfer different viewpoint into arm base coordinate frame. Filters remove the background and noise in the merged seedling point clouds. 3D-to-2D projection and x-axis pixels density distribution method is used to segment the stem and leaves. Finally, separated leaves are fitted with 3D curves for parameter measurement. In testing experiment, 60 maize plants at early growth stage (V2~V5) were sampled by this platform.