AUTHOR=Wang Shubo , Han Yu , Chen Jian , He Xiongkui , Zhang Zichao , Liu Xuzan , Zhang Kai 

TITLE=Weed Density Extraction Based on Few-Shot Learning Through UAV Remote Sensing RGB and Multispectral Images in Ecological Irrigation Area

JOURNAL=Frontiers in Plant Science

VOLUME=Volume 12 - 2021

YEAR=2022

URL=https://www.frontiersin.org/journals/plant-science/articles/10.3389/fpls.2021.735230

DOI=10.3389/fpls.2021.735230

ISSN=1664-462X

ABSTRACT=With current development of ecological irrigation area, higher-level of detection and control categories are required for weeds. In this paper, an improved transfer neural network based on bionic optimization to detect weed density and crop growth is proposed by using the pre-trained AlexNet network for transfer learning. Because the learning rate of new addition layer is difficult to tune to the best, the weight and bias learning rate of the newly added fully connected layer is set by using Particle Swarm Optimization (PSO) and Bat Algorithm (BA) to find the optimal combination on the small data set. By collecting RGB and 5-band multi-spectral of 3 kinds of weeds and 3 kinds of crops as data sets, through cutting, rotating and other operations, data is transported to the CNN. Finally, 6 kinds of classifications are implemented. At the same time, in order to realize the learning of neural network with small sample and high efficiency, a self-constructed CNN based on model-agnostic meta-learning (MAML) is proposed, and its accuracy is verified in the test set. The neural networks optimized by two bionic optimization algorithms are compared with the self-constructed CNN based on MAML and HOG+SVM. The experimental results show that the combination of learning rate through BA is the best. The accuracy can reach 99.39% by RGB images, 99.53% by multi-spectral images, and 96.02% by 6-shot small sample. The purpose of the classification proposed in this paper is to calculate the growth of various plants (including weeds and crops) in the farmland. Then this paper proposes a plant density calculation formula and algorithm. By experimenting in different farmlands, various plant density can be accurately calculated. It provides a basis for the application of variable pesticides in herbicide pollution mitigation. Finally, it promotes a good cycle of ecological irrigation district.