Exploring the Response Mechanism of Remote Sensing Images in Monitoring Fixed Assets Investment Project in Terms of Building Detection

Abstract

Fixed assets investment is a driving factor in facilitating urbanization and economic growth. The local governments have spent a lot of budgets on the construction of fixed assets. However, such investment lacks a scientific and objective mechanism to supervise the construction process continuously to guarantee on-time delivery. Owing to the development of remote sensing technology, the availability of high spatial resolution images makes it possible to visualize the construction process continuously. By synthesizing the amount of fixed assets investment, we can build a reasonable monitoring system to supervise the fixed assets projects mutually in both terms of visual construction and statistical money spent. However, as far as we know, there is not much work exploring the methods in monitoring fixed assets investment yet. We collected the continuous investment records of fourteen fixed assets projects from the year 2015 to 2017 and the corresponding GaoFen satellite images in nine-time nodes. Semantic segmentation deep learning technology is applied to detect buildings from the high spatial resolution images. The monitoring system is built by regression between the ratio of investment and the ratio of building area at nine-time nodes. Compared with the regression model from the ratio of investment and that of ground truth building area, our model achieves an RMSE of 0.0136 in the test samples. It indicates the strong potential applicability of remote sensing images in supervising the reasonability of the construction process of fixed assets and the investment allocation.