Termite identification method based on ResNet neural network

• Main Authors: BAO QIUMEI, GUO JINMIN, TANG XIAOSONG, MA JUN, SHI LEI, ZHA RONGRUI, QIU XIAODI, FAN MU, ZHU DIE, ZHANG ZHIWEI, LUO RONGFU, LUO HANG, YIN CHUAN, SONG BOXU, MA YUNHUA
• Format: Patent
• Language: Chinese; English
• Published: 12.04.2024
• Subjects: CALCULATING; COMPUTER SYSTEMS BASED ON SPECIFIC COMPUTATIONAL MODELS; COMPUTING; COUNTING; PHYSICS

The invention provides a termite identification method based on a ResNet neural network. The method comprises the following steps: acquiring a termite image data set established by termite images of various termite types; performing image enhancement processing on the termite image data set to obtain an enhanced target termite image data set; classifying the target termite image data set into a training set, a verification set and a test set according to a preset proportion in a manner that the numbers of termite images of various termite types are equal; an initial Resnet neural network model is built; inputting a training set and a verification set in the target termite image data set into a Resnet neural network model for training to obtain a network model for identifying termite types; and inputting the test set in the target termite image data set into the network model, and outputting various termite types of the test set, so that the ResNet neural network model is adopted to detect the termite types, a