Two-Stage CNN-Based Wood Log Recognition

The proof of origin of logs is becoming increasingly important. In the context of Industry 4.0 and to combat illegal logging there is an increasing motivation to track each individual log. This work presents a two-stage convolutional neural network (CNN) based approach for wood log tracing based on digital log end images. First, the log cross section is segmented from the background by applying a CNN-based segmentation method using the Mask R-CNN framework. In the second step, wood log recognition is applied using CNNs that are trained on the segmented wood log images using the triplet loss function. Our proposed two-stage CNN-based approach achieves Equal Error Rates between 0.6 and 3.4% on the six employed wood log image data sets and clearly outperforms previous approaches for image based wood log recognition.