基于显著特征和全局特征融合的植物识别方法  

Plant recognition method based on fusion of salient feature and global feature

作　　者：聂聪颖 唐晶磊[1] NIE Cong-ying;TANG Jing-lei(College of Information Engineering,Northwest Agriculture and Forestry University,Yangling 712100,China)

机构地区：[1]西北农林科技大学信息工程学院,陕西杨凌712100

出　　处：《计算机工程与设计》2021年第4期976-984,共9页Computer Engineering and Design

基　　金：国家863高技术研究发展计划子任务基金项目(2013AA10230402);陕西省重点研发计划基金项目(2019NY-170)。

摘　　要：针对自然界中植物数据规模大且分布不平衡导致的识别困难的问题,提出一种基于显著特征和全局特征融合的植物识别方法。通过多层特征融合方法改进VGG19网络以提取植物的全局特征,并对全局特征进行显著特征提取得到图像的显著特征,将显著特征和全局特征进行自适应加权特征融合得到融合特征,对融合特征进行分类识别。在PlantCLEF2016数据集上的实验结果表明,该方法的准确率可达到81%,验证了其在大规模且分布不平衡的植物识别中可行有效。To solve the problem of difficult recognition caused by the large scale and imbalanced distribution of plant data in nature,a plant recognition method based on fusion of salient feature and global feature was proposed.The global feature of the image was extracted using the multi-layer feature fusion method,and salient target extraction was performed on global feature to obtain salient feature of the image.The adaptive weighted feature fusion strategy was used to fuse the salient feature and the global feature.The fusion features were identified in groups.Testing on the PlantCLEF2016 dataset,the results show that the accuracy of the proposed method can reach 81%,which verifies that it is feasible and effective in the large scale and imbalanced distribution situations of plant identification.

关 键 词：植物识别 特征融合 显著特征 大规模数据集 样本不平衡 

分 类 号：TP391.4[自动化与计算机技术—计算机应用技术]