研究目的
Investigating the use of a mobile terrestrial laser scanner (MTLS) for detecting and localizing Fuji apples in an orchard environment to improve crop value and management.
研究成果
The study demonstrates that LiDAR-based technology, particularly its reflectance information, is effective for detecting and localizing Fuji apples in orchard environments. The developed algorithm achieved high detection rates comparable to RGB-based systems, with the added advantages of providing direct 3D fruit location information and being unaffected by sunlight variations. Future work should explore the method's applicability to other fruit varieties and larger datasets.
研究不足
The study's limitations include the small dataset size, which may affect the generalizability of the algorithm to other fruit varieties or species. Additionally, the computational cost increases with the number of points processed, and the method's effectiveness at different laser wavelengths or under varying environmental conditions was not explored.
1：Experimental Design and Method Selection:
The study used a mobile terrestrial laser scanner (MTLS) composed of a Velodyne VLP-16 LiDAR sensor synchronized with an RTK-GNSS satellite navigation receiver to generate 3D point clouds of Fuji apple trees. A reflectance analysis was performed to distinguish apples from other tree elements based on their higher reflectance at 905 nm wavelength. A four-step fruit detection algorithm was developed for apple detection and localization.
2：Sample Selection and Data Sources:
The experiment was conducted in a commercial apple orchard in Agramunt, Catalonia, Spain, focusing on Fuji apple trees. Three trees were analyzed, with a total of 139, 145, and 139 apples manually counted in the field, respectively.
3：List of Experimental Equipment and Materials:
The equipment included a Velodyne VLP-16 LiDAR sensor, an RTK-GNSS satellite navigation receiver (GPS1200+ by Leica Geosystems AG), and a rugged laptop for data processing.
4：Experimental Procedures and Operational Workflow:
The MTLS was moved along a rectilinear trajectory parallel to the tree row at a distance of 2.4 m, scanning the trees from both sides to generate a complete 3D model. The point cloud was then segmented, and a fruit detection algorithm was applied to identify and localize apples.
5：4 m, scanning the trees from both sides to generate a complete 3D model. The point cloud was then segmented, and a fruit detection algorithm was applied to identify and localize apples.
Data Analysis Methods:
5. Data Analysis Methods: The performance of the detection algorithm was evaluated using localization and identification success rates, false detection rates, and F1-score. The reflectance of tree elements was analyzed to distinguish apples from leaves and trunks.
独家科研数据包，助您复现前沿成果，加速创新突破
获取完整内容
