Research On Corn Remote Sensing Estimation Model Based On Neural Network

Crop output is important economic information for a country or region,and the changes in crop output are closely related to people’s lives and social stability.Therefore,it is a very important task for modern society to accurately and effectively predict the yield of crops.As a new type of technology,remote sensing technology has the advantages of short observation period and wide observation range,and can estimate the yield of large areas of crops.At present,the use of remote sensing methods for crop yield estimation has become a mainstream technical means,but this method still has many shortcomings.First of all,due to the influence of terrain and climate,in the practice of remote sensing interpretation,it is found that the quality of remote sensing images obtained is generally low,and they are obscured by a more cloudy layer.In addition,the use of BP neural network combined with remote sensing data for yield estimation in remote sensing yield estimation is easy to fall into a local optimal solution in the learning process.The existence of these problems restricts the development of remote sensing crop yield estimation research.In response to the above-mentioned problems and situations,this article relies on the National Natural Science Foundation of China Project "Research on Intelligent Decision-Making Methods for Corn Precise Operations Based on Multi-source Heterogeneous Big Data",with the theme of remote sensing corn yield estimation from two aspects: improving remote sensing data quality and remote sensing yield estimation accuracy research work.The research area selected the main corn planting area in Jilin Province—Nong’ an County,Changchun City;the research method adopted the popular Conditional Generative Adversarial Net-work(CGAN)in deep learning and the most widely used BP nerve in machine learning.The internet.The main contents of the research are as follows:(1)Constructing a data set of corn yield test.Collect and sort high-scoring remote sensing images and corn production data related to the research area in the National Science and Natural Science Foundation of China and Spark Program projects.At the same time,obtain MODIS remote sensing images and meteorological data in the research area from the NASA website.After preprocessing the data,we use it as the sample data for the production estimation experiment in this article.(2)Propose an improved CGAN network cloud removal method for remote sensing images.For the traditional CGAN network method for remote sensing image cloud removal processing,there is a single feature extraction problem.Combining the idea of multi-scale fusion,the structure and loss function of the CGAN network are improved,and then the optimized CGAN network and other cloud removal methods are simulated and tested.Analyzing the results,a cloud removal method more sui Tab for remote sensing images is obtained.(3)Propose a remote sensing yield estimation method based on optimized BP neural network.Aiming at the problem of long training time and low prediction accuracy of BP neural network in predicting future corn output,particle swarm algorithm is used to optimize the parameters of traditional BP neural network,and a PSO-BP network is established to improve the performance and prediction of the yield estimation model.Accuracy.At the same time,the production estimation results of the PSO-BP neural network and the actual production data are compared and analyzed to verify the effectiveness of the method.(4)Construct a corn yield estimation model based on optimized CGAN and PSO-BP neural network.According to the characteristics of the estimated yield data,the idea of combining the CGAN network and the PSO-BP neural network is determined: First,the optimized CGAN network is used to remove the cloud processing of the remote sensing image,and then the vegetation index such as NDVI and EVI is extracted from the image after the cloud removal,and Use it together with meteorological data as the data input of PSO-BP neural network,and finally train PSO-BP into the network to predict corn yield.By comprehensively analyzing the results of the estimation results and the influencing factors of the model performance,it can be determined that the corn yield estimation model based on neural network established in this paper can better predict the yield of corn,and provide a reliable theoretical basis for farmers to scientifically and rationalize the farming.The increase in income is instructive.