文章信息

標(biāo)題：CNN feature based graph convolutional network for weed and crop?recognition in smart farming

期刊：《 Computers and Electronics in Agriculture》

第一單位：山東農(nóng)業(yè)大學(xué)

在線日期：2020-05-13

Highlights

1.提出了一種基于圖像的半監(jiān)督學(xué)習(xí)方法用于雜草和作物識(shí)別；

2.在四個(gè)不同的雜草數(shù)據(jù)集上進(jìn)行了性能評(píng)估，準(zhǔn)確率高達(dá)98.93%，優(yōu)于傳統(tǒng)的CNN方法；

3.該方法可用于類(lèi)似的識(shí)別任務(wù)。

摘要

除草是提高作物產(chǎn)量的有效方法。準(zhǔn)確可靠的雜草識(shí)別是精準(zhǔn)農(nóng)業(yè)實(shí)現(xiàn)高精度定點(diǎn)除草的前提。為了提高雜草和農(nóng)作物識(shí)別的準(zhǔn)確率，提出了一種基于CNN特征的圖像卷積網(wǎng)絡(luò)(GCN)識(shí)別方法。基于提取的雜草CNN特征及其歐氏距離，構(gòu)建了GCN圖。在半監(jiān)督學(xué)習(xí)的基礎(chǔ)上，GCN圖通過(guò)利用已標(biāo)記和未標(biāo)記的圖像特征來(lái)豐富模型，測(cè)試樣本通過(guò)在圖上進(jìn)行傳播來(lái)從已標(biāo)記的雜草數(shù)據(jù)中獲取標(biāo)簽信息。GCN-ResNet-101方法在4個(gè)不同的雜草數(shù)據(jù)集上的識(shí)別率分別達(dá)到97.80%、99.37%、98.93%和96.51%，優(yōu)于目前最先進(jìn)的方法(AlexNet、VGG16和ResNet-101)。此外，該方法的運(yùn)行時(shí)間也滿足了田間雜草控制的實(shí)時(shí)性要求。本文提出的基于CNN特征的GCN方法有利于在有限的標(biāo)簽數(shù)據(jù)下進(jìn)行多類(lèi)農(nóng)作物和雜草的識(shí)別，在處理類(lèi)似的農(nóng)業(yè)識(shí)別任務(wù)中具有應(yīng)用潛力。此外，所使用的數(shù)據(jù)集和源代碼是公開(kāi)的，以便于在田間雜草識(shí)別方面的研究。

圖3.?基于CNN特征的GCN用于雜草和作物識(shí)別流程

Abstract

Weeding is an effective way to increase crop yields. Reliable and accurate weed recognition is a prerequisite for achieving high-precision site-specific weed control in precision agriculture. To improve weed and crop recognition accuracy, a CNN feature based graph convolutional network (GCN) based approach is proposed. A GCN graph was constructed based on extracted weed CNN features and their Euclidean distances. Based on the semi-supervised learning, the GCN graph enriched the model by exploiting labeled and unlabeled image features, and testing samples obtain label information from labeled weed data by performing propagation over the graph. The proposed GCN-ResNet-101 approach achieved 97.80%, 99.37%, 98.93% and 96.51% recognition accuracies on four different weed datasets respectively, which outperformed the state-of-the-art methods (AlexNet, VGG16 and ResNet-101). Additionally, the runtime of the proposed approach also satisfies the real-time requirement of field weed control. The proposed CNN feature based GCN approach is favorable for multi-class crops and weeds recognition with limited labeled data, which is a promising approach in dealing with similar agricultural recognition tasks. Furthermore, the used datasets and source code are publicly available to facilitate the research in the recognition of field weeds.

總結(jié)

以上是生活随笔為你收集整理的gcn在图像上的应用_每日摘要|基于CNN 特征的图像卷积网络识别杂草和作物的全部?jī)?nèi)容，希望文章能夠幫你解決所遇到的問(wèn)題。

如果覺(jué)得生活随笔網(wǎng)站內(nèi)容還不錯(cuò)，歡迎將生活随笔推薦給好友。