Yang Tao, Wei Jingjing, Xiao Yongjun, Wang Shuyang, Tan Jingxuan, Niu Yupeng, Duan Xuliang, Pan Fei, Pu Haibo
College of Information Engineering, Sichuan Agricultural University, Ya'an, China.
School of Physics and Optoelectronic Engineering, Nanjing University of Information Science and Technology, Jiangsu, China.
Front Plant Sci. 2024 Oct 22;15:1423238. doi: 10.3389/fpls.2024.1423238. eCollection 2024.
Zanthoxylum bungeanum Maxim is an economically significant crop in Asia, but large-scale cultivation is often threatened by frequent diseases, leading to significant yield declines. Deep learning-based methods for crop disease recognition have emerged as a vital research area in agriculture.
This paper presents a novel model, LT-DeepLab, for the semantic segmentation of leaf spot (folium macula), rust, frost damage (gelu damnum), and diseased leaves and trunks in complex field environments. The proposed model enhances DeepLabV3+ with an innovative Fission Depth Separable with CRCC Atrous Spatial Pyramid Pooling module, which reduces the structural parameters of Atrous Spatial Pyramid Pooling module and improves cross-scale extraction capability. Incorporating Criss-Cross Attention with the Convolutional Block Attention Module provides a complementary boost to channel feature extraction. Additionally, deformable convolution enhances low-dimensional features, and a Fully Convolutional Network auxiliary header is integrated to optimize the network and enhance model accuracy without increasing parameter count.
LT-DeepLab improves the mean Intersection over Union (mIoU) by 3.59%, the mean Pixel Accuracy (mPA) by 2.16%, and the Overall Accuracy (OA) by 0.94% compared to the baseline DeepLabV3+. It also reduces computational demands by 11.11% and decreases the parameter count by 16.82%.
These results indicate that LT-DeepLab demonstrates excellent disease segmentation capabilities in complex field environments while maintaining high computational efficiency, offering a promising solution for improving crop disease management efficiency.
花椒是亚洲一种具有重要经济意义的作物，但大规模种植常受到频繁病害的威胁，导致产量大幅下降。基于深度学习的作物病害识别方法已成为农业领域的一个重要研究方向。
本文提出了一种新型模型LT-DeepLab，用于在复杂田间环境中对叶斑病（叶黄斑）、锈病、冻害（冻害）以及患病叶片和树干进行语义分割。该模型通过一个创新的带有CRCC空洞空间金字塔池化的裂变深度可分离模块对DeepLabV3+进行了增强，该模块减少了空洞空间金字塔池化模块的结构参数并提高了跨尺度提取能力。将十字交叉注意力与卷积块注意力模块相结合，为通道特征提取提供了互补的增强作用。此外，可变形卷积增强了低维特征，并集成了全卷积网络辅助头以优化网络并提高模型精度，同时不增加参数数量。
与基线DeepLabV3+相比，LT-DeepLab的平均交并比（mIoU）提高了3.59%，平均像素准确率（mPA）提高了2.16%，总体准确率（OA）提高了0.94%。它还将计算需求降低了11.11%，并将参数数量减少了16.82%。
这些结果表明，LT-DeepLab在复杂田间环境中展现出了出色的病害分割能力，同时保持了较高的计算效率，为提高作物病害管理效率提供了一个有前景的解决方案。
