Automated Classification of Selected Philippine Wood Species Using Image Analysis and Artificial Neural Networks

Abstract

This research aimed to address the need of the wood-based sector for a straightforward, rapid, and reliable wood identification tool. This sector includes agencies like the Department of Environment and Natural Resources, wood processing plants, and state universities and colleges. A model using artificial neural networks was developed to automatically perform image- based identification of 20 selected Philippine wood species. It banks on a progressive database containing numerous macroscopic transverse section images taken from authentic samples of the species included in this study. The model has an F1 score of 87.9%. A system usability survey (SUS) was performed to assess the effectiveness of the web application by deploying it to stakeholders who are engaged in wood identification. The SUS results showed that majority of the respondents rated the web application as either good or excellent. An average of 75.6 SUS score or a grade of “B” (good and acceptable) was obtained from the responses received. All 27 respondents indicated that they would recommend the application to other users. For future directions, inclusion of additional species for identification is recommended, given the fact that there are hundreds of species in the Philippines. This will strengthen the capability of the application to have a more precise and accurate wood identification result. Furthermore, the creation of a mobile application and an offline version of this wood identification app will be taken into consideration.