建筑物变化检测在城市环境监测、土地规划管理和违章违规建筑识别等应用中具有重要作用。针对传统孪生神经网络在影像变化检测中存在的检测边界与实际边界吻合度低的问题，本文结合面向对象图像分析技术，提出一种基于面向对象孪生神经网络（Obj-SiamNet）的高分辨率遥感影像变化检测方法，利用模糊集理论自动融合多尺度下的变化检测结果，并通过生成对抗网络实现训练样本迁移。该方法应用在高分二号和高分七号高分辨率卫星影像中，并与基于时空自注意力的变化检测模型（STANet）、视觉变化检测网络（ChangeNet）和孪生UNet神经网络模型（Siam-NestedUNet）进行比较。结果表明：（1）融合面向对象多尺度分割的检测结果较单一尺度分割的检测结果，召回率最高提升32%，F1指数最高提升25%，全局总体误差（GTC）最高降低7%；（2）在样本数量有限的情况下，通过生成对抗网络进行样本迁移，与未使用样本迁移前的检测结果相比，召回率最高提升16%，F1指数最高提升14%，GTC降低了9%；（3）Obj-SiamNet方法较其他变化检测方法，整体检测精度得到提升，F1指数最高提升23%，GTC最高降低9%。该方法有效提高了建筑物变化检测在几何和属性方面的精度，并能有效利用开放地理数据集，降低了模型训练样本制作成本，提升了检测效率和适用性。

（Objective）Building change detection is essential to many applications, such as monitoring of urban areas, land use management, and illegal building detection. It has been seen as an effective means to detect building changes from remote sensing images. （Method）This paper proposes an object-based Siamese neural network, labelled as Obj-SiamNet, to detect building changes from high-resolution remote sensing images. We combine the advantages of object-based image analysis methods and Siamese neural networks to improve the geometric accuracies of detected boundaries. Moreover, we implement the Obj-SiamNet at multiple segmentation levels, and construct a set of fuzzy measures to fuse the obtained results at multi-levels automatically. Furthermore, we use generative adversarial methods to generate target-like training samples from publicly available datasets to construct a relatively sufficient training dataset for the Obj-SiamNet model. Finally, we applied the proposed method into three high-resolution remote sensing datasets, i.e., a GF2 image-pair in Fuzhou City, and a GF2 image-pair in Pucheng County, and a GF2-GF7 image pair in Quanzhou City, respectively. We also compared the proposed method with three other existing methods, i.e., STANet, ChangeNet, and Siam-NestedUNet. （Result）Experimental results show that the proposed method performed better than three other methods in terms of detection accuracy. More specifically, (1) Compared with the detection results from single-scale segmentation, the detection results from multi-scale increased the recall rate by up to 32%, the F1-Score increased by up to 25%, and the global total classification error (GTC) reduced by up to 7%; (2) When a limited number of samples was available, the adopted generative adversarial network (GAN) was able to generate effective target-like samples for diverting samples. Compared with the detection without using GAN-generated samples, the proposed detection increased the recall rate by up to 16%, F1-Score by up to 14%, and reduced GTC by 9%; (3) Compared with other change detection methods, the proposed method improved the detection accuracies significantly, namely, the F1-Score increased by up to 23%, and GTC reduced by up to 9%. Moreover, the boundaries of the detected changes by the proposed method have a high consistency with that of ground truth. （Conclusion）We conclude that the proposed Obj-SiamNet method has a high potential for building change detection from high-resolution remote sensing images.