Subject: Precision agriculture, Image Processing, Unsupervised Learning
Year: 2015
Type: Proceedings
Title: Unsupervised weed detection in spinach seedling organic farms
Author: Lameski, Petre
Author: Zdravevski, Eftim
Author: Kulakov, Andrea
Author: Gjorgjevikj, Dejan
Abstract: Weed removal during the early phase of seedling development is a very important process in agriculture. It helps the useful plants to sprout quickly and use most of the soil’s organic materials for their own development. The increasing number of human population in the world increases the amount of food that needs to be produced thus the automation of the process of plant based food production is required. In this paper we present an unsupervised approach towards automated weed detection in spinach seedling farms. The images are taken under natural conditions and their green regions are segmented to detect the plants in the images. After that, image descriptors are generated for each plant segment and unsupervised clustering is performed to separate the weeds from the spinach seedlings. The results of the unsupervised learning are compared with the results obtained with supervised learning on the same data. The conclusions are presented in the paper.
Publisher:
Relation: Proceedings of the 24th International Electrotechnical and Computer Science Conference ERK
Identifier: oai:repository.ukim.mk:20.500.12188/21233
Identifier: http://hdl.handle.net/20.500.12188/21233
