Remote sensing image building change detection method fusing UNet + + and residual network

The invention discloses a remote sensing image building change detection method fusing UNet + + and a residual network. The method comprises the following steps: 1, preparing a data set; 2, building a model and training; and 3, detecting changes. A residual network structure is fused into a UNet + + network model, a Res-UNet + + model is established, the model is based on the UNet + + network, a residual convolution unit is introduced, the feature extraction capability of the network is improved, gradient disappearance is avoided, and the building change detection precision is further improved. The experimental results show that, compared with the change detection methods such as UNet and UNet++. The invention has higher overall accuracy and Kappa coefficient. 本发明公开了融合UNet++和残差网络的遥感图像建筑物变化检测方法，包括以下步骤：第一步：准备数据集；第二步：搭建模型及训练；第三步：变化检测。本发明将残差网络结构融合到UNet++网络模型中，建立Res-UNet++模型，该模型以UNet++网络为基础，再引入残差卷积单元，提高网络的特征提取能力，避免梯度消失，进一步提高建筑物变化检测的精度，实验结果表明，与UNet和UNet++等变化检测方法相比，本发明具有更高的整体准确率和Kappa系数。