基于深度学习的人脸识别技术以数据为驱动,对输入图像的质量要求较高。在铁路刷脸进/出站场景下,为滤除因各种因素导致的成像异常的人脸图像,提升人脸识别精度,文章研究人脸图像正常的特征分布,通过知识迁移,提出无须针对异常样本建模...基于深度学习的人脸识别技术以数据为驱动,对输入图像的质量要求较高。在铁路刷脸进/出站场景下,为滤除因各种因素导致的成像异常的人脸图像,提升人脸识别精度,文章研究人脸图像正常的特征分布,通过知识迁移,提出无须针对异常样本建模的人脸图像异常检测算法。理想情况下,该算法对人脸图像异常检测的ROC曲线下面积(AUROC,Aera Under Receiver Operating Characteristic)可达到0.979。实验结果表明,该算法在计算精度与运行成本的组合上具有较高的自由度,可实现不同场景、硬件条件下的算法适配,为优化旅客人脸识别的输入环节,提高各场景下的旅客人脸识别率提供了技术支撑。展开更多
高光谱异常变化检测能够从多时相高光谱遥感图像中寻找到数量稀少、与整体背景变化趋势不同、难以发现且令人感兴趣的异常变化。数据集规模较小、存在噪声干扰以及线性预测模型存在局限性等问题,极大地降低了传统高光谱异常变化检测方...高光谱异常变化检测能够从多时相高光谱遥感图像中寻找到数量稀少、与整体背景变化趋势不同、难以发现且令人感兴趣的异常变化。数据集规模较小、存在噪声干扰以及线性预测模型存在局限性等问题,极大地降低了传统高光谱异常变化检测方法的检测性能。目前,自编码器已被成功地应用于高光谱异常变化检测。然而,单个自编码器在处理多时相高光谱图像时,仅关注图像的重构质量,在获取瓶颈特征时往往忽略了图像中复杂的光谱变化信息。为了解决该问题,提出了一种基于双空间共轭自编码器的多时相高光谱异常变化检测(Multi-temporal Hyperspectral Anomaly Change Detection Based on Dual Space Conjugate Autoencoder,DSCAE)方法。所提方法包含两个共轭的自编码器,即它们从不同方向构造各自的潜在特征。在该方法的训练过程中,首先,两幅不同时刻的高光谱图像经过各自的编码器分别获得相应的潜在空间特征表示,再分别经过各自的解码器获得另一时刻的预测图像;其次,在样本空间和潜在空间中施加不同的约束条件,并在两个空间中最小化相应的损失函数;最后,两幅输入图像经过共轭自编码器后获得各自的异常损失图,对所得的两幅异常损失图采用取小运算得到最终的异常变化强度图,以便在减小输入图像间背景光谱差异的同时突出异常变化。在高光谱异常变化检测基准数据集上的实验结果表明,与10种相关方法相比,DSCAE展现了更优的检测性能。展开更多
For anomaly detection,anomalies existing in the background will affect the detection performance.Accordingly,a background refinement method based on the local density is proposed to remove the anomalies from thebackgr...For anomaly detection,anomalies existing in the background will affect the detection performance.Accordingly,a background refinement method based on the local density is proposed to remove the anomalies from thebackground.In this work,the local density is measured by its spectral neighbors through a certain radius which is obtained by calculating the mean median of the distance matrix.Further,a two-step segmentation strategy is designed.The first segmentation step divides the original background into two subsets,a large subset composed by background pixels and a small subset containing both background pixels and anomalies.The second segmentation step employing Otsu method with an aim to obtain a discrimination threshold is conducted on the small subset.Then the pixels whose local densities are lower than the threshold are removed.Finally,to validate the effectiveness of the proposed method,it combines Reed-Xiaoli detector and collaborative-representation-based detector to detect anomalies.Experiments are conducted on two real hyperspectral datasets.Results show that the proposed method achieves better detection performance.展开更多
文摘基于深度学习的人脸识别技术以数据为驱动,对输入图像的质量要求较高。在铁路刷脸进/出站场景下,为滤除因各种因素导致的成像异常的人脸图像,提升人脸识别精度,文章研究人脸图像正常的特征分布,通过知识迁移,提出无须针对异常样本建模的人脸图像异常检测算法。理想情况下,该算法对人脸图像异常检测的ROC曲线下面积(AUROC,Aera Under Receiver Operating Characteristic)可达到0.979。实验结果表明,该算法在计算精度与运行成本的组合上具有较高的自由度,可实现不同场景、硬件条件下的算法适配,为优化旅客人脸识别的输入环节,提高各场景下的旅客人脸识别率提供了技术支撑。
文摘高光谱异常变化检测能够从多时相高光谱遥感图像中寻找到数量稀少、与整体背景变化趋势不同、难以发现且令人感兴趣的异常变化。数据集规模较小、存在噪声干扰以及线性预测模型存在局限性等问题,极大地降低了传统高光谱异常变化检测方法的检测性能。目前,自编码器已被成功地应用于高光谱异常变化检测。然而,单个自编码器在处理多时相高光谱图像时,仅关注图像的重构质量,在获取瓶颈特征时往往忽略了图像中复杂的光谱变化信息。为了解决该问题,提出了一种基于双空间共轭自编码器的多时相高光谱异常变化检测(Multi-temporal Hyperspectral Anomaly Change Detection Based on Dual Space Conjugate Autoencoder,DSCAE)方法。所提方法包含两个共轭的自编码器,即它们从不同方向构造各自的潜在特征。在该方法的训练过程中,首先,两幅不同时刻的高光谱图像经过各自的编码器分别获得相应的潜在空间特征表示,再分别经过各自的解码器获得另一时刻的预测图像;其次,在样本空间和潜在空间中施加不同的约束条件,并在两个空间中最小化相应的损失函数;最后,两幅输入图像经过共轭自编码器后获得各自的异常损失图,对所得的两幅异常损失图采用取小运算得到最终的异常变化强度图,以便在减小输入图像间背景光谱差异的同时突出异常变化。在高光谱异常变化检测基准数据集上的实验结果表明,与10种相关方法相比,DSCAE展现了更优的检测性能。
基金Projects(61405041,61571145)supported by the National Natural Science Foundation of ChinaProject(ZD201216)supported by the Key Program of Heilongjiang Natural Science Foundation,China+1 种基金Project(RC2013XK009003)supported by Program Excellent Academic Leaders of Harbin,ChinaProject(HEUCF1508)supported by the Fundamental Research Funds for the Central Universities,China
文摘For anomaly detection,anomalies existing in the background will affect the detection performance.Accordingly,a background refinement method based on the local density is proposed to remove the anomalies from thebackground.In this work,the local density is measured by its spectral neighbors through a certain radius which is obtained by calculating the mean median of the distance matrix.Further,a two-step segmentation strategy is designed.The first segmentation step divides the original background into two subsets,a large subset composed by background pixels and a small subset containing both background pixels and anomalies.The second segmentation step employing Otsu method with an aim to obtain a discrimination threshold is conducted on the small subset.Then the pixels whose local densities are lower than the threshold are removed.Finally,to validate the effectiveness of the proposed method,it combines Reed-Xiaoli detector and collaborative-representation-based detector to detect anomalies.Experiments are conducted on two real hyperspectral datasets.Results show that the proposed method achieves better detection performance.
