| Article ID | Journal | Published Year | Pages | File Type |
|---|---|---|---|---|
| 1711401 | Biosystems Engineering | 2012 | 9 Pages |
The transnational project “3D Mosaic” deals with the optimisation of water and fertiliser efficiency in orchards. Detection of the canopy coverage at tree level provides information about the growth capacity of the tree and enables estimation of the possible yield or the influence of reduced water supply in an orchard. Detection must be performed in an automated mode that may be achieved by means of two optical approaches: NIR image analysis, with the calculation of leaf coverage within the image versus non-covered area, and counting the number of laser-scanner (LiDAR) hits per tree. The present study, conducted in an experimental orchard of 180 plum trees, aimed to evaluate and compare these methods using a vertical top-down viewing direction for the sensors. Image analysis showed a higher susceptibility to the sensor mounting height and tilting movements of the carrier vehicle than did the LiDAR measurements. However, on uniform terrain, a Pearson correlation of 0.917 between the systems could be achieved. Both techniques were compared with the manually counted number of leaves per tree for the entire orchard and with the estimated total leaf area for 30 strategically distributed trees. Due to different shapes of the tree crown, the comparison with the leaf numbers yielded lower Pearson correlations for the pollinator cultivar (0.703 with LiDAR, 0.668 with camera) than for the productive trees (0.805 with LiDAR, 0.832 with camera). Comparison of the sensors with the estimated leaf areas yielded correlation coefficients of 0.867 with the laser scanner and 0.788 with image analysis.
► Strong correlation between 2D image and LiDAR for vertically projected plum canopy. ► LiDAR is less sensitive to mounting height and tilting movements than camera sensor. ► Correlation with leaves per tree was greater for more open than more compact habit. ► Correlation with leaf area was 0.867 for LiDAR, 0.788 for image analysis.
