结合加权核范数与3D全变分的目标检测

Object detection with weighted nuclear norm and 3D total variation

针对现实复杂场景使目标检测精确度下降的问题,提出了一种结合加权核范数与3D全变分的目标检测模型。基于扩展的鲁棒主成分分析模型,首先将视频分解为低秩静态背景、稀疏平滑前景和稀疏动态背景,利用加权核范数对背景进行低秩约束,考虑了不同奇异值对秩函数的影响;为加强前景的时空连续性,利用3D-TV来约束运动目标,有效抑制了复杂背景的干扰作用。实验表明,所提算法检测运动目标的准确率较高,能有效抑制复杂背景的干扰作用。

In view of the fact that the accuracy of object detection decreases in complex scenes,a new object detection model based on weighted nuclear norm and 3D Total Variation(3D-TV)is proposed.Based on extended Robust Principal Component Analysis(RPCA),the data are decomposed into low-rank static background,sparse smooth foreground and sparse dynamic background.The weighted nuclear norm is applied to constrain the background,which considers the influence of different singular values on the rank function.In order to enhance the spatio-temporal continuity of the foreground,3D-TV is used to constrain the moving foreground,which effectively suppresses the disturbance of complex background.The experimental results demonstrate that the proposed algorithm has higher accuracy in detecting objects and can effectively suppress the disturbance of complex background.