Tree Detection With Low-Cost Three-Dimensional Sensors for Autonomous Navigation in Orchards

Abstract

This paper deals with autonomous farming and with the autonomous navigation of an agricultural robot in orchards. These latter are typical semi-structured environments where the dense canopy prevents from using GPS signal and embedded sensors are often preferred to localize the vehicle. To move safely in such environments, it is necessary to provide the robot the ability of detecting and localizing trees. This paper focuses on this problem. It presents a low cost but efficient vision-based system allowing to detect accurately, quickly and robustly the trees. It is made of four stereo cameras which provide a point cloud characterizing the environment. The key idea is to find the tree trunks by detecting their shadows which are materialized by concavities in the obtained point cloud. In this way, branches and leaves are not taken into account, improving the detection robustness and therefore the navigation strategy. The method has been implemented using ROS and validated using data sequences taken in several different orchards. The obtained results definitely validate the approach and its performances show that the processing time (around 1ms) is sufficiently short for the data to be used at the control level. A comparison with other approaches from the literature is also provided.

Image-based tree detection for autonomous navigation in orchards

“…To detect the trees present in the robot environment and compute the position of their trunks, we rely on an algorithm that processes in real-time the range component of the point cloud [22]. The method consists in detecting the empty spaces (or shadows), present in the point cloud.…”

“…4, these shadows indicate the presence of trees. Thus, the algorithm first looks for the tree shadows (orange triangles) and then computes their origins (red circles) [22]. This approach offers a high recall rate, i.e., almost all the trees are detected, but it suffers from a fairly low precision rate, i.e., many elements are confused with trees.…”

Row‐sensing templates: A generic 3D sensor‐based approach to robot localization with respect to orchard row centerlines

“…Lyu et al (2018) also proposed a method to detect the boundaries between trunks and the ground and used a naive Bayesian classifier for the free space centerline detection. Durand‐Petiteville et al (2018) presented a stereo vision‐based method to find tree trunks by detecting their “shadows,” that is, concavities in the range component of the obtained point cloud.…”

Detecting and Mapping Trees in Unstructured Environments with a Stereo Camera and Pseudo-Lidar

“…Both of these pipelines use lidar sensing for mapping and tree detection. Finally, Durand-Petiteville et al [9] present a method for detecting tree trunks in orchards using stereo cameras, by reasoning about the shadows that tree trunks cast within the stereo point cloud. This method is computationally efficient and demonstrates that trees can be effectively recognized in stereo point clouds, however since it is designed specifically for usage in orchards, it depends on a number of heuristic assumptions about the shapes and sizes of the trees.…”