Estimation Of Large Area Crop Area Using Multi Scale Remote Sensing Data With Spatial Sampling Method

Results of statistical work in China has played an important role, for the healthy operation of the national economy. However, with the increase of people’s demand for statistical work, the traditional statistical methods have been difficult to meet people’s requirements. In the agricultural production, it is very important to master the information of the cultivated land area in time. Therefore, it is necessary to improve the current statistical survey method and provide a quick, accurate and comprehensive statistical information for the government decision.More and more research and practice show that under the support of remote sensing technology, spatial sampling technology is applied more and more widely based on classical statistical sampling principle combined with spatial statistics theory in agricultural remote sensing monitoring. Spatial sampling method was adopted for the large area crop area monitoring. For example, “Large Area Crop Inventory and Experiment(LACIE)”,“Agriculture and Resources Inventory Surveys Through Aerospace Remote Sensing(AGRISTARS)” and others all use the method of area sampling frame, the monitoring agricultural resources(MARS) unit mission by EU take the stratified sampling method. China has launched a study on a series of research,Huang Huai Hai Plain wheat yield estimation by remote sensing, winter wheat yield remote sensing monitoring in the six provinces of north China,rice yield estimation in southern China and so on. At the same time, how to construct a suitable space sampling method for the study area is the key problem to improve the efficiency.This paper takes the Xinjiang Construction Corps as the research area, proposes the spatial sampling design scheme combing the 3S technology based on divisional and hierarchic method and uses the Multi-temporal remote sensing image to monitor the cotton planting area of Xinjiang Construction Corps. In this paper, through the introduction of the mean and coefficient of variation to evaluate, a lot number of repeated experiments results show that the proposed method is effective and feasible. In the first stage, we make the results of the cotton planting area of the Medium-Resolution image as the true value compared with the deductive cotton area. The precision of the deductive cotton area is more than 95%.The second phase space sampling efficiency evaluation is based on the measured data of the secondary sample, using the weighted variable particle swarm optimization BP neural network to classify the primary sample. The results show that the accuracy of each district is better than 93%, the coefficient of variation is less than 5%, the quality of the thrust is superior. which provides a powerful data support for the relevant policies of the government.