Height estimation for blueberry bushes using LiDAR based on a field robotic platform

Abstract

<abstract> <b><sc>Abstract.</sc></b> Rapid advancement of plant genome research demands field-based high-throughput phenotyping systems. Currently, manual measurement of phenotypic traits such as height in the field is time-consuming and laborious. In this paper, a robotic platform was developed to measure the height of blueberry bushes in the field. A Light Detection And Ranging (LiDAR) was mounted on the platform and was set to scan the bushes from bottom-up view. When the robotic platform was operated in the field, the LIDAR system automatically collected point cloud data of the plants. The height profile was derived after capturing the highest point within each scanning frame, then a sliding window algorithm was applied to extract the height of each bush. Two rows of blueberry bushes (20 samples in each row with varying heights) in the farm were used to test and validate the performance of the proposed system. The two rows were with different height levels. Experimental results showed that the proposed system performed well on the validation data set. An R² of 0.91 and an RMSE of 88.73mm were achieved for all samples. Bush height could be accurately measured using point cloud from either left-side scan or right-side scan. The size of the bushes was a factor influencing the proposed system performance, i.e., an increase of bush height led to a decrease in the accuracy of height measurements.