【摘要】 农作物实景自动监测系统具有自动、非接触、非破坏性等优点,是传统农业气象观测的补充。开展图像质量控制是合理使用农作物实景自动监测系统资料的基础。利用郑州、泰安和固城三地的历史农作物实景图像资料,设计了基于颜色特征参数检测和基于暗通道先验直方图检测的图像数据质量控制方法。通过对2010—2012年三年夏玉米和冬小麦等不同天气条件下得到的农作物实景自动观测资料进行质量控制与应用效果检验。结果表明:两种质量控制方法均可判断出农作物实景自动监测系统中图像观测资料的异常数据;基于颜色特征参数检测方法可有效识别出像素缺失图像,准确率达100%;基于暗通道先验直方图检测方法能有效识别出污染图像,平均准确率为95.7%,平均召回率为87.5%。该质量控制方法可减小模型估算值与观测数据之间的误差,目前已应用于省级农业自动观测业务系统。更多还原

【Abstract】 The automatic monitoring system for crops has the advantages of automatic,non-contact and non-destructive,and is a useful supplement to traditional agrometeorological observations.The charge coupled device(CCD) sensor protects the cover from water,dust,fog-haze,rain and snow,etc.,which may cause observation data error of automatic crop monitoring system.Therefore,the quality control is the basis for the rational usage of the automatic crop monitoring system.Based on the historical crop image data of Zhengzhou,Tai’an and Gucheng,this paper designs a quality control method for image data based on color feature parameter detection and dark channel prior histogram detection.According to different weather conditions,we carried out quality control on the summer maize and winter wheat image observation data in 2010—2012,and the effects were tested.The results show that both types of inspection methods can determine the anomaly data of image observation data in the automatic crop monitoring system.The color feature parameter detection method can effectively identify the missing image of pixels,and the accuracy rate can reach 100%.The proposed method based on the histogram of dark channel can effectively identify the contaminated image,with the average precision being 95.7% and recall average 87.5%.This quality control method can reduce the error between the model estimate and the observed data.At present,this method has been applied to automatic agricultural meteorology observation oprational software.更多还原