Identification of honey bee (Apis mellifera) larvae in the hive with faster R-CNN for royal jelly production

Abstract

Royal jelly is actively used in healthcare products, healthy nutrition, cosmetics industry, strengthening the immune system, treatment of cancer, and many other diseases and new studies are being conducted on its usability in new areas. However, the production of such a useful product is technical and laborious. The most time-consuming process in production is larva grafting performed by hand. For this, a tool that can detect and graft the ideal sized larvae should be developed. The aim of this study, as the first step of such a tool, is to detect the larvae, which are ideal for the production of royal jelly. In the study, initially, a camera setup that can take clear photographs of honeycomb cells was prepared. With this setup, honeycomb photographs were taken containing larvae of different sizes. Later, the larvae with ideal size in the photographs were labelled and the convolutional neural network was trained. Finally, honeycomb cells and centre points were identified with Hough circle, and the locations of the larvae according to the honeycomb cell were determined. In conclusion, a system that can successfully identify ideal sized larvae and their locations to be used in the production process for royal jelly was created.