Deep Learning-Based Pine Nematode Trees’ Identification Using Multispectral and Visible UAV Imagery

Pine wilt disease (PWD) has become increasingly serious recently and causes great damage to the world’s pine forest resources. The use of unmanned aerial vehicle (UAV)-based remote sensing helps identify nematode trees in time a feasible effective approach precisely monitor PWD infection. However, rapid high-accuracy detection not been well established complex terrain environment. To this end, deep learning-based tree identification method is proposed by fusing visible multispectral imagery. A UAV equipped with camera was used obtain imagery, where imagery includes six bands, i.e., red, green, blue, near-infrared, red edge 750 nm. Two vegetation indexes, NDVI (Normalized Difference Vegetation Index) NDRE Red Edge are extracted as typical feature according reflectance infected different spectral bands. YOLOv5 (You Only Look Once v5)-based algorithm adopted optimized from aspects realize high speed accuracy. e.g., GhostNet reduce number model parameters improve speed; module combining CBAM (Convolutional Block Attention Module) CA (Coordinate Attention) mechanism designed extraction for small-scale trees; Transformer BiFPN (Bidirectional Feature Pyramid Network) structure applied fusion capability. experiments show that [email protected] improved 98.7%, precision 98.1%, recall 97.3%, average single 0.067 s, size 46.69 MB. All these metrics outperform other comparison methods. Therefore, can achieve fast accurate trees, providing technical support control epidemic.