Wood Cross-Section Classification Model Based on Lightweight Convolutional Neural Network.

In the context of informatization and intelligence, wood identification technology is also developing in the direction of intelligent and efficiency. In order to solve the problems of long working hours, heavy workload, and human interference in traditional wood identification, in this study, the problem of wood species identification was transformed into a multi-classification problem, and a wood species identification model based on convolutional neural network was developed. The classification of wood cross-section images achieved the effect of tree species identification. In this paper, the cross-section images of eight wood species were taken as the research object, and the images were collected to construct and expand the cross-section data set of tree species. Based on the lightweight convolutional neural network algorithm in deep learning and the inverted residual structure, the lightweight tree species detection model was improved for training and testing. The results showed that the MobileNetV3 network model in the convolutional neural network can accurately and efficiently classify wood cross-section images and then identify wood species. Among them, the improved MobileNetV3 model had certain comprehensive advantages, with the classification accuracy improved by 0.43 % when the number of parameters and the amount of calculation were similar. The overall accuracy of wood cross-section image classification can reach 98.24%. [ABSTRACT FROM AUTHOR]