基于无人机的田间地膜识别算法研究

为了满足基于机器视觉的田间地膜智能识别要求,对垂直拍摄的6叶期单垄单行烟田图像进行研究,分析图像中地膜、烟苗和土壤的RGB和HSV分量的直方图,发现B分量和V分量采用阈值分割法能分割出地膜。对地膜图像进行手动阈值分割、迭代阈值分割、大律法分割和基于遗传算法的最大熵值法等分割算法比较,发现迭代阈值分割算法目标识别率最高,为71%,且处理过程用时较短。通过对无人机不同飞行高度下地膜识别情况分析,发现50 m飞行高度下的地膜识别率最高,为80.06%,地膜识别面积为246.83 m²,地膜覆盖率为30.12%。该研究为残膜的智能识别提供方法和参考。

Research on recognition algorithm of field mulch film based on unmanned aerial vehicle

In order to meet the requirements of intelligent recognition of field mulch film based on machine vision, the images were taken vertically by unmanned aerial vehicle (UAV) to the tobacco seedlings with mulch film in the field. The component color information histograms based on RGB and HSV color model of mulch film, tobacco seedlings and soil were analyzed. The color information of B component and V component was able to identify mulch film from background with the threshold segmentation method. The manual threshold segmentation, iterative threshold segmentation, large-law segmentation and maximum entropy method based on genetic algorithm were compared. Results showed that the iterative threshold segmentation algorithm had the highest target recognition rate of 71%, and the processing time was the shortest. The mulch film recognition rates under different flight heights was analyzed and figured out. It was found that the film recognition rate at 50 m flight height was the highest as 80.06%, the film recognition area was 246.83 m², and the film coverage rate was 30.12%. This research provided a method and reference for intelligent identification of mulch film and residual membrane with crop based on UAV.