Recent advances in the application of deep learning methods to forestry

This paper provides an overview and analysis of the basic theory of deep learning (DL), and specifically, a number of important algorithms were compared and analyzed. The article reviewed and analyzed the main applications of DL methods in forestry including surface quality evaluation of sawn timber, forest resource survey, tree species identification, wood moisture content prediction, the specific application of forestry information text classification, etc. Through comprehensive analysis and review, it was found that: (1) DL method has been widely used in the surface quality evaluation of sawn timber, and the research field mainly uses convolutional neural network (CNN) DL algorithms to carry out research on surface evaluation of sawn timber, and the YOLOv4, YOLOv5m algorithm achieves near real-time target detection and recognition. (2) Establishing a suitable remote sensing image recognition method for forest resources based on DL is a method with great application value in the fields of the future forest resource investigation, statistics of forest vegetation coverage, and monitoring and analysis of plant growth status. (3) The tree species recognition method based on DL effectively avoids the disadvantages of other methods that require image preprocessing for tree images, which leads to cumbersome operation process, low efficiency, and large workload. (4) The DL method provides a quick and efficient prediction method for the prediction of wood moisture content. Moreover, the application of the DL method to the classification of forestry information text provides a new solution to the classification of forestry information text. At the end of the article, a summary of the whole paper is given, and the future development trends of applications of DL to forestry: the field of high-end forestry equipment research, microscopic research in forestry science, and smart forestry are predicted.