| Wheat is one of the three major food crops in China.It plays an important role in agricultural production.Phenotypic information is an important parameter in the dynamic monitoring of crop growth,which is the basis for revealing crop growth patterns and environmental changes.Therefore,the acquisition of wheat phenotypic information has become a research hotspot in current precision agriculture.To efficiently obtain and analyze wheat phenotypic information is still a challenge.In recent years,as a supplement to high-altitude remote sensing,UAV remote sensing has become an important means of low-altitude remote sensing information acquisition.Due to its flexible access and fast timeliness,it has been increasingly used to obtain crop phenotypic information.So,the phenotypic information extraction methods for wheat UAV remote sensing images were studied.The remote sensing images of wheat were obtained by UAV in 2017-2019 including jointing period,heading stage,filling stage and mature stage in Guohe experimental base of Lujiang County,Anhui Province,which were taken as research objects.The main research contents are as follows:(1)The methods based on de-noising for wheat remote sensing image were studied.The mean filtering,median filtering and adaptive median filtering were used to de-noising the wheat remote sensing images with different growth periods,different heights and different varieties,and the image quality of wheat remote sensing images after de-noising by three different filtering methods was compared and analyzed qualitatively and quantitatively.It is concluded that the adaptive median filtering method can not only reduce the image noise,but also protect the edge information of the image.(2)The methods based on fusion for wheat remote sensing images were studied.The direct averaging,wavelet transform and weighted average were used to fuse the wheat remote sensing images with different growth periods,different heights and different varieties.The results of experiment show that there are differences in stitching marks and brightness in the overlapping area of the fusion image.Therefore,the weighted average method was improved.The improved weighted average method was used to fuse the wheat remote sensing images of different growth periods,different heights and different varieties.After fusion,the stitching traces in the overlapping areas of the image are not only eliminated,but also the brightness of the image tends to be consistent.The improvement of image quality after fusion by the improved weighted average method and other three fusion methods is compared and analyzed qualitatively and quantitatively.The average gradient,information entropy,mean and standard deviation of proposed method were improved by 9.5%,4.1%,2.9%,and 8.4%compared with the weighted average method,respectively.(3)A method based on UAV for monitoring the nitrogen content of wheat leaves was proposed.The monitoring process of nitrogen content of wheat leaves at different growth stages was complicated and destructive.So a method for rapid prediction of nitrogen content of wheat leaves at different growth stages by using UAV remote sensing images was proposed.The color and texture features of wheat phenotypic information were extracted from the spliced and fused wheat remote sensing images.The color and texture features of wheat phenotypic information extracted from splicing and fusion wheat remote sensing images were used as input of the support vector machine regression model to predict the nitrogen content of wheat leaves at different growth stages.The accuracy of the prediction model of leaf nitrogen content(LNC)of wheat leaves at different growth periods based on texture features,color features and texture&color features was compared and analyzed.It was concluded that it was more accurate to predict nitrogen content of wheat leaves by using texture&color features model. |
