当前位置: X-MOL 学术Precision Agric. › 论文详情
Our official English website, www.x-mol.net, welcomes your feedback! (Note: you will need to create a separate account there.)
Evaluation of weed impact on wheat biomass by combining visible imagery with a plant growth model: towards new non-destructive indicators for weed competition
Precision Agriculture ( IF 5.4 ) Pub Date : 2021-01-02 , DOI: 10.1007/s11119-020-09776-6
Christelle Gée , Emmanuel Denimal , Josselyn Merienne , Annabelle Larmure

To evaluate the impact of weeds on crops, precise identification and early prediction are required. This paper presents two new non-destructive indicators deduced from visible images: weed pressure (WP) and wheat growth status (WGS). They are based on the fractional vegetation cover (FVC) obtained from digital vegetation maps (crop vs. weeds) in a wheat field. FVC was determined for both plants with a Matthews Correlation Coefficient of 0.86 using machine learning classification [support vector machine-radial basis function (SVM-RBF)] combined with Bag of Visual Words technique. It was compared to destructive measurements of above-ground biomass (BM) and leaf area index (LAI). Since the coefficient of determination between FVC and BM is very good for wheat crop (r 2 = 0.93), FVC is used to feed a growth model based on the Monteith equation. Replacing the standard approach by the image approach in the initialization of the model had no impact on the simulated BM values. WP characterized weed pressure, namely the FVCw/FVCc ratio and it quantified the crop–weed competition. The results show that up to the tillering stage, it could substitute for the BMw/BMc ratio resulting from a destructive approach. The second indicator, WGS assessed crop health through the monitoring of biomass production. It compared the theoretical wheat biomass simulated under non-stressed conditions, BM simulated , to the actual biomass, BM observed . The impact of weed on crop was evaluated by combining the results of these two indicators. This simple and fast method based on proximal detection data offers promising results in agroecological cropping systems, where high responsiveness is a major challenge for site-specific weed management.

中文翻译:

通过将可见图像与植物生长模型相结合来评估杂草对小麦生物量的影响:实现杂草竞争的新非破坏性指标

为了评估杂草对作物的影响,需要精确识别和早期预测。本文介绍了从可见图像中推导出的两个新的非破坏性指标:杂草压力 (WP) 和小麦生长状况 (WGS)。它们基于从麦田中的数字植被图(作物与杂草)获得的部分植被覆盖率 (FVC)。使用机器学习分类 [支持向量机-径向基函数 (SVM-RBF)] 结合视觉词袋技术,确定了两种植物的 FVC,马修斯相关系数为 0.86。将其与地上生物量 (BM) 和叶面积指数 (LAI) 的破坏性测量进行了比较。由于 FVC 和 BM 之间的决定系数非常适合小麦作物 (r 2 = 0.93),因此使用 FVC 为基于 Monteith 方程的生长模型提供数据。在模型的初始化中用图像方法代替标准方法对模拟的 BM 值没有影响。WP 表征了杂草压力,即 FVCw/FVCc 比率,它量化了作物 - 杂草的竞争。结果表明,直到分蘖阶段,它可以替代破坏性方法产生的 BMw/BMc 比率。第二个指标,WGS 通过监测生物量生产来评估作物健康。它将在非胁迫条件下模拟的理论小麦生物量(BM 模拟)与实际生物量(BM 观察)进行了比较。综合这两个指标的结果来评价杂草对作物的影响。这种基于近端检测数据的简单快速的方法在农业生态种植系统中提供了有希望的结果,
更新日期:2021-01-02
down
wechat
bug