Development of mobile-based application for practical wood identification

Abstract

Wood Identification is an important task, especially for wood anatomist. The identification process is required in many fields, such as custom ports, forest survey, and wood industries. However, expertise-in wood identification is limited due to a few wood anatomists. To enable wood identification can be performed in a broader area, a mobile-based application is proposed to realise these purposes. The paper presents the development of a mobile-based application for wood identification. A dataset of Indonesian commercial wood images of cross-section surface was collected from Xylarium Bogoriense, The Ministry of Environment and Forestry of The Republic of Indonesia. The images were collected using a smartphone camera at optical magnification level around 200 times and a minimum resolution of 12 megapixels. The collected images were then used to develop a deep learning-based algorithm to classify wood species. These images are considered as the training datasets, whereas testing dataset is collected in another session. The final model obtained through the training process is then stored in a cloud-based server at Research Center for Informatics, Indonesian Institute of Sciences. A mobile application, namely AIKO stands of Indonesian words-Alat Identifikasi Kayu Otomatis (Automatic Tool for Wood Identification), is developed to enable image acquisition and images transferring to the server. AIKO application sends the acquired image to the trained identification model in the server. The output of the identification model is the species name of the observed images. Besides species/botanical name, AIKO also provides information on trade names, specific gravity and density, durability class, strength class, commercial timber classification, and recommended utilization.