Study On Real-time Detection Of NDVI In Corn Canopy And Intelligent Fertilization Zoning Method

Topdressing is an agricultural production process that increases corn yield and quality in the early stage of corn.The method of topdressing with average fertilization amount.The average topdressing amount is insufficient for fertilizer supply in areas with low soil fertility,but excessive topdressing is caused in areas with high soil fertility.Due to significant differences in soil fertility at different locations.At the same time,problems such as low fertilizer utilization rate and environmental pollution have seriously restricted the sustainable development of agriculture.The fertilization zone was established through the variable fertilization model and the spatial distribution of emotional information to effectively improve fertilizer utilization.An intelligent,real-time corn topdressing partitioning method was proposed based on fertilization partition idea and NDVI.The main content is to obtain the NDVI of corn canopy in the middle cultivating period by vehicle-mounted canopy spectroscopy sensor,and study its relationship with corn yield and nitrogen content,and establish a fertilization model based on NDVI variable.On this basis,the NDVI spatial variability law is analyzed to determine a reasonable sampling method,and the intelligent zoning algorithm based on big data is discussed to establish an intelligent method of topdressing and zoning,and the variable fertilizer application machine is controlled in real time.The main research contents and results are as follows.(1)The best working parameters of the vehicle plant canopy spectral sensor(Greenseeker)were analyzed to obtain the Canopy NDVI of Demeia 1 and Demeia 3 maize varieties.The variable topdressing model of different maize varieties of NDVI was established based on the nitrogen fertilizer recommendation algorithm to analyze the relationship between NDVI and yield.The results show that the light intensity has no significant effect on the Greenseeker detection accuracy,and the best detection height is 60-100 cm.Compared with the conventional topdressing treatment,the variable topdressing reduced the amount of topdressing by 31.5%(Demeta 3)and 30.5%(Demeta 1).(2)Errors were analyzed based on differential NDVI time series data by establishing ce lls with different plant spacing and nitrogen application rates.On this basis,the NDVI m ean,maize plant height and nitrogen content estimation accuracy of dense plant spacing were used as evaluation indicators to compare different error elimination algorithms.The results show that gross error is the main reason for the accuracy of NDVI data.The diff erence of NDVI mean value and NDVI mean value of dense plant spacing after box lin e treatment under different plant spacing was the smallest,which were 9.2% and 22.3%,respectively.At the same time,the estimation accuracy of maize plant height and nitrog en content is higher than other algorithms.The box line method effectively reduces the i nterference of the error on the corn canopy NDVI.(3)The accuracy of the large-area NDVI data acquired by Greenseeker was discussed by The variability of the difference sequence of corn canopy NDVI time series data obtained by comparing UAV remote sensing with Greenseeker.On this basis,the spatial variability analysis of NDVI is carried out by using regional variable theory to determine the spatial variability and scale dependence of NDVI at small scales,and the distribution law of corn growth potential and the optimal sampling interval are obtained.The results show that there is no significant difference in the trend of NDVI time series data acquired by different canopy spectral sensors(sig>0.05),and the accuracy of NDVI time series data acquired by Greenseeker was higher.As the detection distance increases,the NDVI autocorrelation weakens,and the geomorphic factors are the main factors affecting the growth of the corn.The block gold value and the abutment value ratio show a smooth change when the detection interval is 10 meters,and the optimum inspection interval is determined to be 10 meters.(4)The clustering was compared by the clustering evaluation index and the estimation accuracy of the corn growth parameter.Base on NDVI of Greenseeker,which shows that K-mean can establish reasonable partitioning under four clusters.At the same time,the grid density clustering method was used for the dynamic NDVI incremental data.The data itself is used to adjust the parameters of the clustering algorithm,and the parameters are self-adjusted for real-time cluster analysis Under different data quantities,the NDVI distribution histograms in the corresponding partition clusters of the two algorithms are used to match the area operation similarity evaluation index to evaluate the performance of the algorithm.The results show that the similarity of the partition map obtained by K-mean and real-time clustering increases with the amount of detected data,and the highest similarity(0.72)is obtained with the largest data,indicating the idea based on big data.The real-time partitioning method is feasible.