Wood Species Recognition Using Machine Learning

Although a computer vision-based automated wood species detection system is not extensively employed today, it is in high demand in a range of industries. Because each species of wood has distinct features and traits that determine how it is used and implemented. Human experts have traditionally been used nearly entirely in the identifying process, which is based on a variety of features of wood, such as texture, structure, and colour. Even competent scientists have difficulties distinguishing between different species of wood found in nature. The most difficult aspect of wood recognition is that some wood species have a texture that is identical to that of others, hence feature extraction is the most important step in the process. So in this work, two statistical feature extraction techniques are used 1) Local Binary Pattern (LBP) 2) Gray Level Co-Occurrence Matrix (GLCM). Features extracted from both methods are combined to see whether they increase classification accuracy or not. Multi-SVM is used for classification. For this study, a new dataset is created which includes the three most common tree species in Indian territory. This study found that integrating LBP and GLCM features enhanced classification accuracy. Proposed system produces a result of 97.2%, which is better than individual approaches.