The continued development of new, high yielding cultivars needed to meet the world’s growing food demands will be aided by improving the technology to rapidly phenotype potential cultivars. High-throughput phenotyping (HTP) is essential to maximize the greatest value of genetics analysis and to better understand the plant biology and physiology in view of a “Feed the World in 2050” theme. Field-based high-throughput phenotyping platform including a LiDAR-based proximal sensing field scout and its data processing system were developed and applied for near-real-time remote sensing of soybean 3D canopy structure variation and plant growth condition among different plant and plot scale in this paper. 

The proximal sensing field scout consists of an extremely accurate distance measurement sensor, called the SICK LiDAR (Light Detection And Ranging) scanner, with distance measurements over a 180 degree area up to 8 meters away among canopy. The GPS sensor and a 6 DOF (Six Degrees of Freedom) Inertia Measurement Unit were mounted on the tractor. The scanning frequency is 1Hz and the resolution will be 10mm. The canopy height and density will be calculated after the canopy 3D structure is reconstructed. The resolution is 0.16m / pixel, which is much smaller than the individual soybean plant size of 0.762m. The ground reference data of each plot, such as canopy height information, were collected during the growing season, and the yield harvested at the end of the season was measured. These ground-based agronomic traits data will be correlated with the ground proximal sensing data. 4 rows of the soybean were proximal sensed and scouted by LiDAR considering the traveling speed and the sensor resolution. The average maximum height of the plant and the canopy coverage were derived, and the plant canopy volume was calculated based on the canopy 3D reconstruction algorithm. Therefore, a novel canopy volume model algorithm was developed to correlate with the yield considering the canopy height and canopy coverage information based on the entire field and each soybean family evaluated. The results shows that the canopy height by proximal sensing of LiDAR has a good correlation with that of field measurement (R²=78.51%) and the canopy volume value has a correlation of 69.34% (R²=69.34%) with soybean yield harvested in 2013. Additionally, some of the families has a positive correlation between the yield harvested and the soybean canopy volume across the whole field and the proposed canopy volume model has better stability and robustness than that of the field height only. It is indicated that proposed proximal sensing system is high efficiency and effective to conduct the field-based high-throughput phenotyping and which could be helpful to better understand the plant biology and growth condition for plant improvement. Commercial pre-harvest yield prediction would likely to be made during the early season based on the proximal sensing data.  Moreover, the proximal sensing approach for high-throughput phenotyping could be applied in QTL and association mapping with crop genetics in the future.