A method of calculating phenotypic traits for soybean canopies based on three-dimensional point cloud

Analysis of soybean phenotypes is a core motivation behind soybean breeding. However, amounts of manual measures are required in obtaining canopy phenotypic traits via traditional methods. Moreover, deficiencies such as time consumption, strong subjectivity, and inaccuracy can be also detected in manual measurement. In order to achieve automatic extraction of phenotypic traits in the research of soybean breeding, a method of acquiring soybean phenotypic traits was proposed on the basis of a Kinect sensor with three soybean varieties (incl. KANGXIAN9, KANGXIAN13, and FUDOU6) as research objects, implementing the calculation of plant height (PH), leaf area index (LAI). Firstly, the canopy image information was acquired vertically to extract canopy data with the registration of color images and depth point cloud data. Secondly, a soybean single plant was segmented from the group canopy using the bounding box method; also, the height of the soybean plant was solved using the distance information; meanwhile, the canopy LAI was calculated with extinction coefficients that were optimized by the beer-lambert law. According to experimental results, determination coefficients R 2 of the calculated value and the measured value of the plant height and the leaf area index of the three soybean varieties are greater than 0.94. It can be seen that calculated results can meet the accuracy requirement of phenotypic traits in soybean breeding. • An automatic method of acquiring soybean phenotypic traits was proposed. • The bounding box method was used to segment single plant from the group canopy. • The plant height, LAI was calculated by distance value and the beer-lambert law. • The R 2 of plant height and LAI were 0.97 and 0.94, respectively.