Abstract: The study on the change patterns of linear features in remote sensing imagery is a very active research field in land use survey worldwide. This research can provide important information and reference in liner feature extraction and area calculation. Statistical regression method was used to analyze 105 random samples collected in northern China (Beijing, Tianjin, and Hebei Province). By comparing the measured width with the appearance of the corresponding liner features in Landsat TM7 image, the general change patterns of linear features in Landsat TM7 image can be obtained. From the analysis, it demonstrates that different liner features, such as road and ditch, exhibit different change patterns due to real difference in width and orientation. Even for the same type of linear features with the same width and orientation, their change patterns on TM image can be rather different. This paper analyzed the major causes contributing to the change patterns.