Development of Cabbage Detection Program for Computing the Vertical Projection Leaf Area in Aerial Images

The vertical projection leaf area (VPA) of cabbage (Brassica oleracea var. capitata) is an important phenotypic trait that is closely related to total leaf area, biomass, and head weight (yield). Developing a high-throughput method to measure individual VPAs of cabbage can assist in monitoring growth conditions in the field and improve the prediction of crop growth. High-throughput phenotyping using drones that can extract phenotypic crop traits from aerial images is widely used. Here, an R program was developed to detect and segment cabbages in images and calculate each cabbage’s VPA. The detection error and precision of the method were evaluated by using orthoimages collected from cabbage fields of NARO, Kannondai, Tsukuba, Ibaraki, Japan. The average detection error was 3.93%–5.40% and the average detection accuracy was 93.03%–96.35%. The accuracy of the VPA generated by the method was sufficient to be used in a cabbage growth prediction model and has the potential to predict yields of individual cabbages.