Deep learning applications for oil palm tree detection and counting
Kuryati Kipli, Salleh Osman, Annie Joseph, Hushairi Zen, Dayang Nur Salmi Dharmiza Awang Salleh, Asrani Lit, Kho Lee Chin
Abstract
Oil palms are one of the essential crops in agricultural productivity for developing countries such as Malaysia and \nother tropical areas. For predicting the yield and production of palm oil, the counting process is often carried out. Manually counting oil palm trees is one of the solutions but it requires a massive labour force, and the result is often inaccurate. To overcome this problem, automated techniques for oil palm detection have been developed. However, the performance of existing automated techniques for oil palm detection deteriorates when the \nplanting layout of the oil palm tree is not well organized. Deep learning applications for oil palm tree detection \nand counting offer a powerful solution to the challenges of precision agriculture, enabling plantations to increase \nproductivity and sustainability while reducing costs and manual labour. Deep structured learning, more generally deep learning is one of the widely used computer vision technology, especially in agricultural engineering. Deep learning method is an essential tool when it comes to monitoring the plantation. Different deep learning networks are utilized for classification tasks towards oil palm trees. In order to promote the use of deep learning in the oil palm industry, this paper main contribution is to provide an understanding of the utilisation of deep learning and its application in oil palm tree counting. The gaps and opportunities for research in oil palm plantations based on deep learning techniques will also be discussed.