Yield Estimation Based On UAV Remote Sensing Monitoring Of Cotton Growth Process Method Study

Timely and accurate estimation of cotton yield is of great practical significance to the field management and production of cotton crops and the improvement of final yield under the background of agricultural modernization.The advantages such as high resolution and fast acquisition speed of UAV image data can be used to realize rapid monitoring of crop planting areas and accurate acquisition of various characteristics of cotton crops,thus realizing reliable monitoring of crop information.In addition,the rapid development and progress of UAV also make up for the low resolution,slow acquisition speed and high acquisition cost of high altitude satellite remote sensing data,so as to be widely used in various fields of agriculture.Based on the time series UAV high-resolution images,this study selected the cotton growing area of Shihezi Reclamation area of Xinjiang Corps as the research object to carry out related experimental research.In this experiment,the time series images of the study area were firstly preprocessed by splicing,correcting,clipping and registration,so as to obtain the complete time series images of the whole study area.Secondly,the 34 planting quilt index was used for comparative analysis.After the segmentation threshold was determined by Otsu method,the seedling image of the study area was segmtioned.After that,the cotton plants were extracted by agricultural toolkit in ENVI5.6.At last,the main information of cotton emergence uniformity,seedling growth and later growth was monitored,and the final yield of the study area was predicted by using all kinds of key information of the whole growth period combined with relevant ground statistical data.Through experiments,the conclusions of this study mainly include:1.Extraction of cotton seedlings.After comparative analysis of various vegetation indices,it was found that the green leaf index was the best for image segmentation.After segmentation of seedling images by vegetation index combined with Otsu method,cotton extraction in the study area was realized by crop counting tool in ENVI5.6 agricultural toolkit.According to statistics,283,410 cotton seedlings were extracted from the whole study area.The repeated tests at different scales showed that the extraction accuracy of cotton in the study area reached 96.2%,and the experimental results were good.2.Monitoring cotton growth in later period.According to the monitoring of cotton emergence uniformity and coverage at seedling stage in the study area,the emergence rate and growth state of the western and southern plots were higher,while the emergence rate and growth state of the northern plots were poor.By using NDVI to monitor the monthly growth of cotton in the study area,the growth rate of cotton in the study area was the fastest from full flowering to full bolling stage.From the seedling stage,the overall growth state of cotton in the northern part of the plot was poor,while that in the western part of the plot was good.3.Cotton yield estimation.According to the correlation analysis with the real yield of the sampling area,the correlation between NDVI and the real yield of the three-stage image was 0.51,0.61 and 0.76,and the correlation between NDVI and the emergence rate and the coverage of the seedling stage was 0.81and 0.46,respectively.In the weighted normalized vegetation index constructed based on time series images,the weights of the images on June 17,July 18 and August 7 were 0.27,0.32 and 0.41 respectively.The cotton yield in the study area was 7244.55kg/hm~2by constructing the yield estimation model.The determination coefficient R~2of the model was 0.84,and the root mean square error RMSE was 0.31.Compared with the real yield,the yield estimation accuracy of the model reached 95.11%.The yield estimation accuracy of the model reaches 94.10%,both of which have achieved good experimental results.