基于无人机影像数据的烟草精准识别方法研究

Research on Tobacco Precise Recognition Method Based on UAV Image Data

为有效减少杂草对烟草可见光影像识别提取的影响,实现烟草的精准识别,利用四旋翼无人机采集贵州山区烟草种植基地的正射影像(digital orthophoto map,DOM),根据地物的光谱特征差异,构建颜色重组差异植被指数(color recombination difference vegetation index,CRDVI),运用最大类方差(OTSU)确定分割阈值,实现烟草识别和提取,并与常见的可见光差异植被指数(visible-band difference vegetation indx,VDVI)、过绿指数(excess green,ExG)和改进型绿红植被指数(improved green-red vegetation index,MGRVI)对比分析。结果表明:CRDVI可有效去除杂草对于烟草识别和提取的影响;在杂草覆盖度较高的地块,CRDVI对烟草识别效果优于VDVI、ExG和MGRVI;CRDVI在3个实验样区中和验证样区中识别和提取的完整性均在89%以上,提取精度较高。研究结果证实CRDVI可快速准确提取烟草并可有效抑制杂草的影响,可为山区特色作物的识别提取提供参考。

In order to effectively reduce the impact of weeds on the identification of visible light images of tobacco and achieve accurate identification of tobacco,quadrotor UAVs were used to collect digital orthophoto map(DOM)at cultivation of tobacco base in Guizhou mountainous areas.According to the spectral characteristics of ground objects,CRDVI(color recombination difference vegetation index)was constructed,and OTSU was used to determine the segmentation threshold to achieve tobacco identification and extraction,and it was compared and analyzed with the common visible and difference vegetation index(VDVI),excess green(ExG),and improved green red vegetation index(MGRVI).The results indicated that CRDVI could effectively remove the impact of weeds on tobacco recognition and extraction;CRDVI outperformed VDVI,ExG,and MGRVI in identifying tobacco in plots with high weed coverage;the integrity of CRDVI recognition and extraction in the 3 experimental and validation sample areas was above 89%,with high extraction accuracy.The experiment confirmed that CRDVI could quickly and accurately extract tobacco and effectively restrain the influence of weeds,which provided a reference for the identification and extraction of characteristic Cash crop in mountain areas.