Study On Weed Distribution And Threat Assessment In Spacious Seedling Wheat Field

Seedling crop growth is highly susceptible to weed pests and diseases.Currently,the rough spraying herbicide way in large scale,not only cause environmental pollution,but also cause pharmaceutical waste and weed resistance.In order to use herbicides rationally and reduce the impacts of weed diseases on seedling growth,we used the seedling field images which captured in the experimental filed of Northwest A&F University.After using digital image processing technology which combined with GIS technology,we carried out data extraction and interpolation analysis of the farmland images,and than obtained a spatial distribution of weed and assessed weed threats.Based on these results,a prescription map of the sprayed drug partition was generated to guide the field variable spraying operation.Here is the main contents and conclusions of this paper:(1)Scheme design and farmland data collection.In order to assess the overall threat level of weeds in the farmland,determine the weed distribution and avoid costly manpower and material resources,the "X" and grid two sampling methods were designed to realize the sampling of the whole farmland data as much as possible.In this paper,in order to obtain the high quality farmland image,precise sample coordinates,and avoid issues like identifying the individual size only,density of weeds,which caused by lowly spectral resolution,the digital camera combined with GPS was used to sample the farmland image data.(2)Farmland image segmentation.In this paper,the green plant and soil background were segmented by color characteristics of color image.Because the single color factor cannot obtain the better gray image,this paper compared the four common color factors(ExR,2G-R-B,ExG-ExR and H component),and used the maximum interclass variance method to realize the binarization segmentation of farmland image.The experimental results showed that the gray scale(ExG-ExR factor)evaluation coefficient e value was 1803.025 and the binary segmentation precision PS value of the H component was 0.9637.At this point,the evaluation value was largest and the gray-scale effect was best.We studied and improved the pixel position histogram method,and than achieved the separation of weed and crop.The experimental results showed that the improved algorithm can better adapt to the skew in the field crop lines.(3)Weed distribution and threat assessment based on GIS.In order to realize the visualization of weed distribution and weed threat assessment,this paper interpolated and evaluated the data extracted by "X" and grid sampling by Kriging interpolation.After that,we conducted weed threat assessment,and established a prescription map of farmland pharmacy spraying based on weed threat visualization.Finally,farmland weeding operations were guided by this method.The experimental results showed that the block gold effect of the two sampling methods was between 25% and 75%,which indicated that the weed density of the two sampling methods hasd moderate spatial autocorrelation.The RMSSE values of the Gaussian model were closest to 1 in the overall interpolation result of different variation function model,It was shown that the Gaussian model had the best fitting effect and the interpolation precision was the highest.In this paper,the study of weed distribution and threat assessment has a certain practical significance for the rational use of herbicides,the protection of the environment and the protection of crop safety.