Preliminary Study On The Rural Population Estimation Method Based On Remote Sensing Image

China is a country with large population, and most of it comes from countryside. It is significant to do the research on the changes of rural population for the social economic and political environment, and many other fields, but the traditional census has problems of long period and high cost. With the rapid development of remote sensing technology, the resolution of remote image is much higher than before, and the application on rural population estimation has been widely applied. At present, most research in this field mainly make city as the study area for population estimates, the study in rural areas is less. In order to solve this problem, this paper, taking Jiexiu as an example, by adopting the object-oriented method do the extraction of residents on remote sensing image, combined with the second land survey results of Jiexiu, based on eCognition7.0, analyzed the correlation between it and the rural population, and establish the model of population estimation based on remote sensing image.The main results are as follows:1、Do the repeated testing to determine the optimal multi-scale segmentation parameters, on true color IKONOS images composed closely to gray image in winter or early spring, the parameters(150,0.2,0.5) can accurately extract the rural residential area, on IKONOS image in summer or early autumn, the parameters(100,0.3,0.5) can finish it well, and extraction accuracy can reach84%.2、Analysis showed that there is a positive correlation between the residential area of the106villages for the model and the its population. Then make population as the dependent variable, the extracted rural residential area as the independent variable, establish a linear regression model of rural population estimation based on remote sensing images, Y=0.007X+2.493,the coefficient of the determination is0.857.3、Using the data of90villages that not involved in modeling to verify the model, the accuracy can reach99.38%, the error is only0.62%. Doing the further analysis get that:when population size is [0,1000], the model estimated results was instability and high volatility, when population size is [1000,3000], the model estimated results was more stable, when the population scale is more than3000, the samples in this interval is less. But the tendency show that the greater the population size, the greater the stability of the model estimation results.4、Combined with the DEM of Jiexiu, analyzed the relationship between terrain and the error, show that the applicability of the model is better than the hilly area in the plains region.The research shows that:the model established in this paper can estimates the rural population at a lower cost, more accurately and quickly. The study has a definite value for population estimation and provides a new idea.