Performance Enhancement of Automatic Wood Classification of Korean Softwood by Ensembles of Convolutional Neural Networks

Abstract

In our previous study, the LeNet3 model successfully classified images from the transverse surfaces of five Korean softwood species (cedar, cypress, Korean pine, Korean red pine, and larch). However, a practical limitation exists in our system stemming from the nature of the training images obtained from the transverse plane of the wood species. In real-world applications, it is necessary to utilize images from the longitudinal surfaces of lumber. Thus, we improved our model by training it with images from the longitudinal and transverse surfaces of lumber. Because the longitudinal surface has complex but less distinguishable features than the transverse surface, the classification performance of the LeNet3 model decreases when we include images from the longitudinal surfaces of the five Korean softwood species. To remedy this situation, we adopt ensemble methods that can enhance the classification performance. Herein, we investigated the use of ensemble models from the LeNet and MiniVGGNet models to automatically classify the transverse and longitudinal surfaces of the five Korean softwoods. Experimentally, the best classification performance was achieved via an ensemble model comprising the LeNet2, LeNet3, and MiniVGGNet4 models trained using input images of $128{\times}128{\times}3pixels$ via the averaging method. The ensemble model showed an F1 score greater than 0.98. The classification performance for the longitudinal surfaces of Korean pine and Korean red pine was significantly improved by the ensemble model compared to individual convolutional neural network models such as LeNet3.

Fig. 1. A confusion matrix and their meaning.

Fig. 2. Comparison of diagonal elements of confusion matrix from LeNet-type and MiniVGGNet-type models for different sizes of input images.

Fig. 3. Normalized diagonal values from confusion matrices by ensemble models for input images of 64 ×64 × 3.

Fig. 4. Normalized diagonal values from confusion matrices by ensemble models for input images of 128 ×128 × 3.

Fig. 5. Confusion matrices and F1 scores of several ensemble models by the averaging method.

Table 1. Class name and its designated index for species and surface combination

Table 2. The architecture of LeNet models.

Table 3. The architecture of MiniVGGNet models.

Table 4. The best results by SNR-like measure from ensemble sets for input image of 64 × 64 × 3.

Table 5. The best results by SNR-like measure from ensemble sets for input image of 128 × 128 × 3.

Table 6. Performance measures of LeNet2-LeNet3-MiniVGGNet4 ensemble model by the averaging method.

Table 7. Performance measures (F1 scores) of the best ensemble model (LeNet2, LeNet3, and MiniVGGNet4) by the averaging method and individual CNN models (LeNet3 and MiniVGGNet3).