Monitoring Of Land-cover Change Around Highway Using Remote Sensing Data

Road construction is restricted by natural factors such as geologic structure, terrain, hydrological condition and productivity. Land development has impacted the environment inevitably by destroying the soil, vegetation and all kinds of landscape.Remote sensing data can be used to monitor the land-cover change around the highways in a macro scale synchronously. This is its advantage. In this study, Land sat Thematic Mapper (TM) data of the Qinghai-Tibet highway of 1986 and 1994 is used to compare the land-cover change. Statistics and artificial intelligence method are combined to heighten the classified precision. This method can provide quantificational gist for road environment issue, road location selection, and landscape design.Recently, along with the developing of artificial neural network theory, neural network technology becomes an effective instrument to deal with remote sensing data classification. In this thesis, two neural network models are used to classify the TM image after Principal Component Analysis: BP neural network and Self-organizing feature map neural network. BP neural network is widely used. Therefore, there are some disadvantages such as low training efficiency and local minimum. The construction and parameter of SOFM are simpler, and classification accuracy is higher. In this thesis, first three Principal Component of TM data is used in image classification. Then Contingency matrix is used to evaluate the classification precision. By comparing 3 kinds of accuracy and Kappa quotient, conclusion is obtained: the classification accuracy of SOFM is higher than BP and MLC; the classification ability of BP is not as good as MLC. Overall accuracy of SOFM is 94.0%, Kappa is 0.9114, and overall accuracy is 14.9% and 9.8% higher than BP and MLC. After comparison SOFM is used to classify image of 1986, the land-cover changes of two year are compared.