SLAM-based incremental convex hull processing approach for treetop volume estimation

• Treetop volume information of a grove is important for agricultural processes.
• LiDAR scanning provides 3D information of the grove with very low error.
• An optimized convex hull algorithm is used to estimate treetop volume based on the LiDAR readings.
• The LiDAR scanning and the convex hull algorithm are sensitive to localization errors of the sensor.
• A SLAM algorithm is implemented to minimize localization errors and to improve treetop volume estimation.

Treetops volume information in groves is a key component of the perception process for improving herbicide management, foliage density observation and a grove’s canopy maturity supervision. In this work, a computational geometry based approach for estimation of the treetops of a grove is implemented and tested. This approach is based on convex hull techniques to estimate the volume of a treetop from 3D raw laser data. The method shown here optimizes both the computational cost associated with the convex hull processing and the volume of stored information, which become crucial for in-field experimentation. Additionally, this work presents an analysis of how the localization of the range sensor used for treetop volume estimation, directly affects the information regarding such treetop. Thus, a mathematical and empirical analysis of treetop volume estimation using a GPS antenna and a SLAM (Simultaneous Localization and Mapping) algorithm is included, showing that the SLAM algorithm provides with a better estimation. The mathematical foundation of the proposal, as well as convergency tests and real-time experimentation results are also shown in this work.