A new alternative for assessing ridging information of potato plants based on an improved benchmark structure from motion

Ridging based on crop height is essential for promoting potato plant growth and increasing potato yield. However, existing crop height estimation methods are greatly affected by the flatness of the digital terrain model (DTM), and estimation results with low leaf area are often unsatisfactory. To quickly and accurately estimate potato plant height in the early growing stage and guide ridge operation, an unmanned aerial vehicle (UAV) visible light system was used to collect RGB images of potato plants shortly after sowing and on the 50th day after planting. A red–green fit index (RGFI) was constructed using a digital orthophoto map (DOM) on the 50th day after planting potato plants and the Gaussian mixture model (GMM) threshold method to classify the experimental plots. The mean value of soil digital surface models (DSM) after classification was taken as DTM, and the benchmark structure from motion (BSfM) was proposed in combination with the structure from motion (SfM) method. The results showed that the BSfM had a higher correlation with the measured plant height than the SfM method, with an R2 value of 0.7807 and RMSE value of 1.8272 under second-order polynomial function fitting, indicating that BSfM could effectively minimise the effect of DTM flatness on plant height estimation. BSfM relies only on DSM distribution with soil and potato plants and can be easily calculated, with great potential as a rapid and cost-effective tool to estimate crop height.