| 引用本文: | 马淳,李世雄,达布希拉图,等.基于数字图像技术的西兰花氮素诊断施肥研究[J].灌溉排水学报,2025,():-. |
| MAChun,LIShixiong,DA Buxilatu,et al.基于数字图像技术的西兰花氮素诊断施肥研究[J].灌溉排水学报,2025,():-. |
|
| 摘要: |
| 【目的】准确地诊断作物氮素营养状况，实现作物精准施肥和氮肥资源合理利用，利用数码图像参数预测西兰花氮素营养指标，明确氮素营养状况评价的最佳色彩参数和方程模型，指导西兰花氮素营养管理。【方法】通过设计五个不同水平的氮素梯度试验，分别为CK(不施用氮肥处理)、N1(施用172.5 kg/hm2氮肥处理)、N2(施用258 kg/hm2氮肥处理)、N3(施用345 kg/hm2氮肥处理)、N4（施用510 kg/hm2氮肥），分析西兰花氮素吸收规律并建立西兰花冠层图像的多元逐步回归模型，在考虑经济可行的前提下，以当地常规施肥和NDVI模型施肥为对照，进行田间诊断施肥验证。【结果】施氮量显著影响西兰花叶片含氮量、产量及冠层色彩参数。西兰花的吸氮量随着施氮量的增加而增加，当施氮量大于258 kg/hm2时生物量不再增加。利用Matlab 2018软件对冠层图像进行图像去躁和灰度化处理，再利用多元逐步回归方法对各颜色参量进行分析，B（蓝光值）与吸氮量之间存在显著正相关关系，R2达到0.863；比颜色指标S/H和S/V值与吸氮量之间也存在着正相关关系，R2分别达到0.697、0.737。通过试验获得施氮量和产量之间的关系，利用多元逐步回归的方法初步构建了基于数码相机拍摄冠层数字化指标（B值、S/H值以及S/V值）和NDVI值（归一化植被指数）的氮肥追施模型。通过一年三点的验证试验发现，利用模型诊断施肥在试验点一的施氮量下降13.1%~17.5%；试验点二下降7.7%~18.3%；试验点三下降12.2%~19.4%。采用TOPSIS多准则决策模型，比较了不同模型施肥条件下的产量和生物量，综合评价不同模型诊断施肥，表现为RGB模型>HSV模型>NDVI模型>常规施肥。【结论】数字图像技术可应用于西兰花氮素营养诊断研究，最佳诊断施肥模型为RGB模型，作为一种新方法在西兰花氮素营养快速无损诊断、精准施肥及减少化肥施用量，在可持续农业中具有较好的应用潜力。 |
| 关键词: 西兰花;NDVI;RGB;HSV;氮素营养诊断 |
| DOI：10.13522/j.cnki.ggps.2025122 |
| 分类号:S-3210 |
| 基金项目:土壤碳源互作调控西兰花GLS代谢的营养机理研究(4246070427) |
|
| Research on Nitrogen Diagnosis and Fertilization Management in Broccoli Using Digital Imaging Technology |
|
MAChun1, LIShixiong2, DA Buxilatu2, ZANGChonghui2, PENG Yuanyang2
|
|
1.Yunnan Agricultural University;2.College of Resources and Environment, Yunnan Agricultural University
|
| Abstract: |
| 【Objective】To accurately diagnose the nitrogen (N) nutritional status of crops, enabling precision fertilization and rational utilization of N fertilizer resources, this study utilized digital image parameters to predict N nutritional indicators in broccoli (Brassica oleracea var. italica). It aimed to identify the optimal color parameters and regression models for evaluating N status, thereby guiding broccoli N nutrition management. 【Methods】A field experiment was conducted with five nitrogen application gradients: CK (no N fertilizer), N1 (172.5 kg/hm2 chemical fertilizer), N2 (258 kg/hm2 chemical fertilizer), N3 (345 kg/hm2 chemical fertilizer), and N4 (510 kg/hm2 chemical fertilizer). Broccoli nitrogen uptake patterns were analyzed, and a multiple stepwise regression model was established based on broccoli canopy images. Considering economic feasibility, field validation of diagnostic fertilization was performed, using local conventional fertilization and NDVI model-based fertilization as controls. 【Results】Nitrogen application rate significantly affected broccoli leaf nitrogen concentration, yield, and canopy color parameters. Broccoli N uptake increased with increasing N application; however, biomass ceased to increase when the application rate exceeded 258 kg N/ha. Canopy images were processed for noise reduction and grayscale conversion using Matlab 2018 software. Multiple stepwise regression analysis revealed a significant positive correlation (R2 = 0.863) between the Blue (B) band value and N uptake. Positive correlations were also found between N uptake and the specific color indices Saturation/Hue ratio (S/H) (R2 = 0.697) and Saturation/Value ratio (S/V) (R2 = 0.737). The relationship between N application rate and yield was established. Subsequently, a topdressing recommendation model was preliminarily constructed using multiple stepwise regression based on digital canopy indices derived from digital camera images (B value, S/H, S/V) and the Normalized Difference Vegetation Index (NDVI). Validation trials conducted across three locations over one growing season demonstrated reductions in N application rates using the model: Site 1: 13.1%–17.5%; Site 2: 7.7%–18.3%; Site 3: 12.2%–19.4%. The Technique for Order of Preference by Similarity to Ideal Solution (TOPSIS) multi-criteria decision model was employed to compare yield and biomass under different model-based fertilization strategies. The comprehensive evaluation ranked the diagnostic fertilization models as follows: RGB model > HSV model > NDVI model > conventional fertilization. 【Conclusion】 Digital image technology can be effectively applied to broccoli nitrogen nutrition diagnosis. The RGB model was identified as the optimal diagnostic fertilization model. This approach represents a novel method for rapid, non-destructive diagnosis of broccoli N status, enabling precision fertilization and reducing fertilizer application. It demonstrates significant potential for application in sustainable agriculture. |
| Key words: Broccoli; NDVI ; RGB; HSV; Nitrogen nutrition diagnosis |
