Wood Type Classification Based on Hybrid Feature Integration with Optimized Bagging Ensemble Approach

Images of oak (Quercus petrea L.), chestnut (Castanea sativa M.) and Scots pine (Pinus sylvestris L.) tree species, which are widely used in Türkiye and around the world, were obtained in this study using mobile devices. The primary objective of this study is to automatically and reliably distinguish these wood species using image processing techniques and statistical classification methods, thereby enabling tree species identification at the genus level. In this context, colour and edge-based features such as HSV (Hue, Saturation, Value), LAB (Lightness, A (green–red), B (blue–yellow)), LBP (Local Binary Pattern) and Sobel (Sobel Edge Detection Operator) were extracted from the images. These features were evaluated using Random Forest, XGBoost, CatBoost, and Extra Trees algorithms to test classification performance. The experimental results show that colour-based features such as HSV and LAB achieved 97.5% accuracy with the Extra trees algorithm, while 100% accuracy was achieved with an optimisation-based bagging ensemble approach using all features together. Achieving such high accuracy on real-world data collected in the field using mobile devices demonstrates that the proposed method can be used as a reliable species identification tool in practical applications.