Plant phenotyping techniques are important to present the performance of a crop and it interaction with the environment. The phenotype information is important for plant breeders to analyze and understand the plant responses from the ambient conditions and the inputs offered for it. However, for conclusive analysis it is necessary a large number of individuals. Thus, phenotyping is the bottleneck of plant breeding, a consequence of the labor intensive and costly nature of the classical phenotyping. Consequently, efficient high throughput phenotyping (HPP) is needed. In this scenario, many studies have evaluated the use of sensors for the development of an efficient HPP. Therefore, the aim of this study was to develop a greenhouse structure for plant phenotyping and to test sensors in order to evaluate the advantages and disadvantages of it for plant phenotyping. A structure with three rails was developed for scanning two vases with soybean plants. A camera T3 Canon, a LMS-200 (LiDAR) and a Kinect version 1 (K1) were used to generate the 3D models of the plants. According to the results, the LiDAR sensor generated the point cloud with the other two sensors. On the other hand, Kinect and T3 RGB cameras are very affected by the ambient light. Moreover, the sun light limits outdoor uses of the K1 sensor because the infrared from the sun amess with the infrared pattern generated by K1 used to measure the depth distances. In terms of processing time consuming, the Structure From Motion 3D reconstruction is the most time consuming. In general, LiDAR creates a robust result but still is a more expensive sensor, K1 is not very suitable for field conditions (sun light exposure) and RGB cameras can be used all conditions but processing is computer intensive.