Change information of linear objects can be detected based on linear feature in multi-temporal remote sensing images. A new linear feature change detection approach is proposed. In this paper, the changed edge points from multi-temporal remote sensing images are grouped into line support regions based on the difference of gradient information. Then, the regions are fitted based on least square and the changed lines are generated. This approach of linear features change detection uses the gradient information of multi-temporal images as the change detection element instead of match the line information, so it decrease the complexity of the change detection. The experimental results in multi-temporal remote sensing show the effective of the proposed approach.