Honey bees play an essential role in the food chain, being responsible for one third of the global food supply due to pollination. Thus, preserving the health of beehives is of paramount environmental and economic importance. Unfortunately, at present a decline in bee populations is reported, attributed to factors such as climate change, environmental disasters, use of pesticides, etc. The SmartBeeKeep (https://smartbeekeep.eu/) research project, co-funded by EU and Greek funds, builds on the latest developments in remote sensing and AI technologies to provide a holistic platform (currently at the integration stage) that offers services addressing different needs of beekeepers and associated researchers, facilitating their work and contributing to the study of biodiversity. Specifically, an automated mapping service was implemented that runs periodically in the back end and uses the freely available multi-temporal and multi spectral Sentinel-2 data to estimate and update information regarding beekeeping flora (including blooming detection), based on state-of-the-art AI models for semantic segmentation. Moreover, a web/mobile mapping app and a mobile (progressive web) app were developed, exploiting modern remote sensing and AI technologies. In particular, the mapping app displays freely available data layers that provide crucial information for beekeepers and enables them to view and edit their own data layers, manually entering information regarding beekeeping flora near their apiaries. On the other hand, the mobile app provides two additional functionalities: a) tools for beehive inspection and management, which allow beekeepers to keep track of honeybee colonies development, applied treatments and/or feeding actions, and b) automated AI-based identification of beekeeping plants from photos captured by the mobile phone. An e-marketplace for beekeeping products as well as additional services towards laboratories performing analyses of beekeeping products as well as the general public are also included. Preliminary results for two variants of the automated mapping procedure based on a new dataset including a beekeeping plant are also presented. The final goal is improve current beekeeping practices, reduce costs, and create a new distribution channel for beekeeping products.
The accelerated advancements in remote sensing technologies and the deployment of satellites offering freely accessible multispectral satellite imagery have facilitated the application of machine learning, particularly deep learning techniques, to tasks such as crop classification, yield estimation, and bloom detection. Additionally, several countries in the European Union have adopted the Land Parcel Identification System (LPIS), that obliges farmers to declare the exact area and crop type of their parcels each year while also making the LPIS data freely accessible to the public. For the purpose of the SmartBeeKeep research project, co-funded by EU and Greek funds, in this work we utilize the above with the objective to combine multispectral and multitemporal satellite data obtained from the Sentinel-2 satellite with the LPIS parcel maps in order to detect and classify the blooming period of the beekeeping plant lavender, with the use of automated deep machine learning methods. The specific plant type was selected as it is exhibits particular interest to the beekeeping community, which is the main focus group of the SmartbeeKeep project. For this task, a dataset was amassed and thoroughly sorted out, that comprises of approximately 15k individual parcels from the area of Southern France between January 2020 to December 2021. For each parcel, a study of its harmonized EVI index was carried out in order to roughly identify its blooming period temporal boundaries and with the help of experts, characteristic parcels for sub-regions of Marseille were selected to create more accurate temporal annotations. Additionally, freely available data from the EU Copernicus DIAS reference service WEkEO were utilized as an initial temporal estimation for the Start-of-Season (SOS) and End-of-Season (EOS) period. Two temporal deep learning methods were evaluated, namely a convolutional and a recurrent and model, so as to establish benchmark results on the created dataset. The dataset will be released upon publication.
Access to the requested content is limited to institutions that have purchased or subscribe to SPIE eBooks.
You are receiving this notice because your organization may not have SPIE eBooks access.*
*Shibboleth/Open Athens users─please
sign in
to access your institution's subscriptions.
To obtain this item, you may purchase the complete book in print or electronic format on
SPIE.org.
INSTITUTIONAL Select your institution to access the SPIE Digital Library.
PERSONAL Sign in with your SPIE account to access your personal subscriptions or to use specific features such as save to my library, sign up for alerts, save searches, etc.