| Autor: |
Pflanz, Michael, Nordmeyer, Henning |
| Jazyk: |
German<br />English |
| Rok vydání: |
2016 |
| Předmět: |
|
| Zdroj: |
Julius-Kühn-Archiv, Iss 452, Pp 241-248 (2016) |
| Druh dokumentu: |
article |
| ISSN: |
1868-9892 |
| DOI: |
10.5073/jka.2016.452.033 |
| Popis: |
Unmanned Aerial Vehicles (UAV) have become omnipresent and adequate tools to generate high-resolution spatial data of agricultural cropland. Their implementation into remote sensing approaches of weeds provides suitable applications for a site-specific herbicide management. In general, an increasingly use of innovative technologies gradually leads from agricultural research into the practical application. This requires an evaluation of possibilities and limits of UAV-based remote sensing procedures. While spectrals from UAVs are being used already for mapping needs of nutrient or water, the image supported weed detection is much more complex and at the moment not relevant in practice. In this regard, there is a lack of weed and crop differentiation through spectral analyses and object-based approaches separate different plants not species-specific or are not adapted to morphologic changes of the growth. Moreover, there is a need for alternative positioning techniques without GPS, as it is required for a precise optical imaging analysis at low altitudes. To evaluate the possibilities and limitations of automated weed identification regarding the optical and sampling requirements, flights were carried out with a hexacopter at an altitude of 5 m over agricultural crop land with variable weed patches. The altitude was controlled by the GPS-autopilot. Images were captured at geo-referenced points and the number of different weed species was simultaneously determined by manually counting. The required optical resolution on the ground was estimated by comparing the number of weeds between image analysis on the PC and with the field rating data. |
| Databáze: |
Directory of Open Access Journals |
| Externí odkaz: |
|
