基于高分辨率图像的多尺度作物分类

Multi-scale Crop Classification Based on High-resolution Images

基于无人机平台获取的地面影像有着较高的空间分辨率,但提供丰富的细节信息的同时,也为农作物分类带来很多“干扰”,尤其是在利用深度模型进行作物识别时,存在边缘信息提取不充分及相似纹理作物误分,导致分类效果欠佳等问题.因此,通过多尺度注意力特征提取的思路构建模型,有效提取边缘信息,提高作物分类精度.所提出的多尺度注意力模型(multi-scale attention network,MSAT)通过多尺度块嵌入获取同一层级不同尺度的作物信息,多尺度特征图被映射为多条序列独立地馈送到因子注意力模块中,增强对农作物上下文信息的关注,提高模型对地块边缘信息的提取,因子注意力模块内置的卷积相对位置编码增强块内部局部信息的建模,提高对相似纹理作物的区分能力,最后通过融合局部特征与全局特征,实现粗细双重信息的提取.在水稻、甘蔗、玉米、香蕉和柑橘5种作物上的分类结果表明,MSAT模型的MIoU(mean intersection over union)和OA(overall accuracy)指标达0.816、98.10%,验证了基于高分辨率图像的精细作物分类方法可行且设备成本低.

The ground images obtained by the unmanned aerial vehicle(UAV)platform have a high spatial resolution,but they also bring a lot of“interference”to crop classification while providing rich details.In particular,when depth models are used for crop recognition,there are problems such as insufficient edge information extraction and misclassification of similarly textured crops,which results in a poor classification effect.Therefore,a model is constructed by the idea of multi-scale attention feature extraction to effectively extract edge information and improve the accuracy of crop classification.The proposed multi-scale attention network(MSAT)obtains crop information on different scales at the same level through multi-scale block embedding.The multi-scale feature map is mapped into multiple sequences that are fed into the factor attention module independently,which enhances the attention to crop contexts and improves the model’s extraction ability of plot edge information.Moreover,the built-in convolutional relative position encoding of the factor attention module enhances the modeling of local information inside the module and the ability to distinguish similarly textured crops.Finally,the thickness information is extracted upon the fusion of local features and global features.The classification results of rice,sugarcane,corn,bananas,and oranges show that the mean intersection over union(MIoU)and overall accuracy(OA)of the MSAT model reach 0.816 and 98.10%,respectively,which verifies that the fine crop classification method based on high-resolution images is feasible,and the equipment cost is low.