基于直方图的手机玉米冠层数字图像氮素诊断方法

Nitrogen Nutrition Diagnosis Method Based on Mobile Phone Image of Summer Maize Canopy

【目的】便捷准确地诊断作物氮素营养状况是实现作物精准施肥和氮肥资源合理利用的关键。近年来应用数码相机等工具进行作物营养诊断的研究受到广泛关注。本研究采用智能手机相机获取玉米冠层图像,建立完善的基于手机相机的氮素营养诊断技术,并比较传统的图像均值方法和直方图方法对氮素营养诊断的可靠性,以探明夏玉米氮素营养诊断的最佳适用模型。【方法】基于田间氮肥用量试验,采用智能手机相机获取夏玉米拔节期冠层图像,提取夏玉米冠层图像的G/R、G/B、NRI[R/(R+G+B)]、NGI[G/(R+G+B)]、NBI[B/(R+G+B)]和(G-R)/(R+G+B)6种颜色指数均值及直方图敏感区间,分别建立冠层图像色彩参数均值模型与直方图模型,分析其与玉米叶片含氮量和产量的关系。利用决定系数(R^(2))、均方根误差(RMSE)、平均绝对百分比误差(MAPE)对比不同指数模型模拟估算玉米叶片含氮量和产量的稳定性和准确性,建立基于手机相机获取夏玉米冠层图像的氮素营养诊断模型。【结果】施氮量显著影响玉米叶片含氮量、产量及冠层图像色调和植被覆盖度。直方图波峰b随叶片含氮量的增加而发生变化,相较于冠层图像色彩参数指数均值方法,指数直方图法适用于不同品种的氮素诊断。色彩参数(G-R)/(R+G+B)直方图可以更好地反映作物覆盖率及整体颜色信息,指数直方图与玉米叶片含氮量和产量也呈现较好的相关性。基于神经网络模型验证数据集精度评价指标,指数直方图模型中玉米叶片含氮量和产量的MAPE值和RMSE值均低于指数均值模型,R^(2)达到0.753,大于指数均值模型。指数直方图模型验证结果MAPE值达到5.80%,RMSE值为0.07,估算精度高,泛化性强。结果表明,冠层图像色彩参数指数直方图在估算叶片含氮量和产量时具有更高精度和更强鲁棒性,能够有效利用玉米叶片覆盖度、颜色等特点,具有较好的稳定性。【结论】利用智能手机相机获取玉米冠层数字图像,结合冠层图像色彩参数直方图方法建立的神经网络模型具有较好的应用效果,提高了估测精度,作为一种新方法在玉米氮素营养快速无损诊断和精准施肥中具有较好的应用潜力。

【Objective】Convenient and accurate diagnosis of crop nitrogen(N)status is the key to achieve precise crop fertilization and rational utilization of N resources.In recent years,the application of digital cameras and other tools in crop nutrition diagnosis has attracted wide attention.In this study,the smart phone cameras were used to obtain maize canopy images,and nitrogen nutrition diagnosis technology based on mobile phone cameras was established and improved.The reliability of traditional image mean method and histogram method for nitrogen nutrition diagnosis was compared to find out the best model for nitrogen nutrition diagnosis of summer maize.【Method】Based on the experiment of N fertilizer amount in the field,the canopy image of summer maize at jointing stage was obtained by smartphone camera.Six color indices,including G/R,G/B,NRI[R/(R+G+B)],NGI[G/(R+G+B)],NBI[B/(R+G+B)]and(G-R)/(R+G+B),were extracted from summer maize canopy images,and the histogram sensitive interval were established,respectively,to analyze their relationship with leaf N content and yield of maize.The determination coefficient(R^(2))and root mean square error(RMSE)were used to determine the relationship between the mean color index model and the histogram model.Mean absolute percentage error(MAPE)was used to simulate and estimate the stability and accuracy of leaf N content and yield in maize compared with different index models.Then,the N nutrition diagnosis model based on mobile phone camera acquisition of summer maize canopy images was established.【Result】N application significantly affected leaf N content,yield,canopy image hue and vegetation coverage of maize.The peak b of the histogram changes with the increase of leaf N content.Compared with the mean color index method in canopy images,the index histogram method was suitable for N diagnosis among different varieties.The color index(G-R)/(R+G+B)histogram could better reflect crop coverage and overall color information.The index histogram also showed a good correlation with leaf N content and yield.Based on the neural network model to validate the accuracy evaluation indicators of the dataset,the MAPE and RMSE values of leaf N content and yield in maize in the exponential histogram model were lower than those in the exponential mean model,and the R^(2) reached 0.753,which was greater than that in the exponential mean model.The validation results of the exponential histogram model showed a MAPE value of 5.80%and an RMSE value of 0.07,indicating high estimation accuracy and strong generalization ability.The results indicated that the color parameter index histogram of canopy images had higher accuracy and stronger robustness in estimating leaf N content and yield,and could effectively utilize the characteristics of maize leaf coverage,color,etc.,with good stability.【Conclusion】Therefore,the neural network model established by using smartphones to obtain digital images of maize canopy and combining them with the color index histogram method of canopy images has good application effects and improves estimation accuracy.As a new method,it has good potential in rapid and non-destructive diagnosis of maize N nutrition and precise fertilization.