Int. Arch. Photogramm. Remote Sens. Spatial Inf. Sci., XLI-B3, 863-870, 2016
https://doi.org/10.5194/isprs-archives-XLI-B3-863-2016
© Author(s) 2016. This work is distributed under
the Creative Commons Attribution 3.0 License.
 
10 Jun 2016
BEESMART – A CROWDSOURCING PROJECT WITH SMARTPHONES
E. Gülch1, S. Uddin1, and B. Willi2 1Hochschule für Technik Stuttgart, Schellingstr 24, 70174 Stuttgart, Germany
2Imkerverein Waiblingen e.V. Waiblingen, Germany
Keywords: Flower recognition, Honey Yield Web Portal, Geo-localization, UAV Abstract. The project Beesmart aims at the derivation of a geolocation yield catalogue for honey bees by using a crowd-sourcing approach with the help of smartphones. A central issue are thus the design of an application (App2bee) for smartphones and the design of a software for flower recognition, which uses sensor information of the smart phone and information about blooming times to recognize and localise flowers. The implemented flower recognition is based on the approach “Minimal-bag-of-visual-Words“. A classification accuracy of about 60-70% can be reached, which is of course affected by the big variety of flowers, by the way on how images are taken and how the image quality and resolution actually are. The classification results are further improved by applying apriori a simple manual segmentation on the touch screen to put the focus in the image on the flower in question. The design and the functionality of the App2Bee are presented followed by details on the communication, database and Web-portal components. In a second part of the project the classification of larger areas of flowers important for honey bees are investigate using a fixed-wing UAV system with two different types of cameras, a RGB digital camera and a NIR digital camera. It is certainly not possible to recognize single flowers, but it could be shown, that larger fields of the same flower, like e.g. Red Clover, can be classified with this approach. With the data available it was also possible to classify bare-ground, roads, low pasture, high pasture as well as mixed pasture. For the high pasture it was possible to automatically identify clusters of flowers, like Yarrow.
Conference paper (PDF, 1614 KB)


Citation: Gülch, E., Uddin, S., and Willi, B.: BEESMART – A CROWDSOURCING PROJECT WITH SMARTPHONES, Int. Arch. Photogramm. Remote Sens. Spatial Inf. Sci., XLI-B3, 863-870, https://doi.org/10.5194/isprs-archives-XLI-B3-863-2016, 2016.

BibTeX EndNote Reference Manager XML