Search

Addressing the spatial and temporal variability of crops for agricultural management requires intensive and periodical information gathering from the crop fields. Unmanned Aerial Vehicle (UAV) photogrammetry is a quick and affordable method for information collecting; it provides spectral and spatial information when required with the added value of Digital Surface Models (DSMs) that reconstruct the crop structure in 3D using 'structure from motion' techniques. In the full process from UAV flights to image analysis, DSM generation is one bottle-neck due to its high processing time. Despite its importance, the optimization of the required forward overlap for saving time in DSM generation has not yet been studied. UAV images were acquired at 50 and 100 m flight altitudes over two olive orchards with the aim of generating DSMs representing the tree crowns. Several DSMs created with different forward laps (in intervals of 5-6% from 58 to 97%) were evaluated in order to determine the optimal generation time according to the accuracy of tree crown measurements computed from each DSM. Based on our results, flying at 100 m altitude and with a 95% forward lap reported the best configuration. From the analysis derived from this configuration, tree volume was estimated with 95% accuracy. In addition, computing time was 85% lower in comparison to the maximum overlap studied (97%). It allowed computing the 3D features of 600 trees in a 3-ha parcel in a highly accurate and quick (a few hours after the UAV flights) manner by using a standard computer. [ABSTRACT FROM AUTHOR]

Perennial ryegrass (Lolium perenne) is a perennial crop used in temperate regions as forage. In L. perenne breeding programs, persistency is an important trait. Poor persistency results in sward degradation and associated yield and nutritive value losses. Breeders assess persistency of accessions using visual scoring in field plots during the 2nd or 3rd growing season. This evaluation system is easy and cheap but is not free from human bias. In this study, the correlation between the scoring done by different breeders was only 0.243. As an alternative we have developed a methodology to assess persistency of L. perenne breeding materials based on vegetation indices (VIs) derived from Unmanned Aerial Vehicle (UAV) imagery. The VIs Excess green (ExG2), Green Leaf Index and Normalized green intensity (GCC) were found to provide consistent results for flights carried out under different light conditions and were validated by ground reference information. The correlation between the VIs and the percentage of ground cover extracted from on-ground imagery was 0.885. To test the implementation of the method we compared the ExG2 value based approach to selection with a visual score based selection methodology as applied by two breeders. The breeding decisions of Breeder A agreed well with decisions based on ExG2 values (74.6%), but those of Breeder B displayed a lower agreement (54.0%). In contrast, agreement between decisions based on different flights was very high (91.6%). The methodology was validated for general applicability. In summary, the results demonstrate that basing persistency selection in L. perenne breeding programs on ExG2 values from UAV imagery is likely to be more objective in comparison to the currently-used visual scoring method. [ABSTRACT FROM AUTHOR]