基于超像素一致显著性的视频运动目标检测算法

Video Moving Object Detection Algorithm Based on Super-pixel Uniform Saliency

针对目前视频运动目标检测中存在受背景影响大,复杂场景下检测效果不佳的问题,提出一种基于超像素一致显著性的视频运动目标的检测方法。首先,将视频图像序列进行超像素分割,在保留目标特征完整性的基础上降低图像后续处理的计算复杂度。之后对视频单帧图像进行显著性检测,得到图像显著目标区域,接下来通过对视频图像序列间显著区域超像素匹配机制对运动目标进行检测。最后,引入视频图像序列间显著性传播的协同判别因子提高对运动目标的判别精度。实验结果表明,所提算法具有较强的鲁棒性,能够处理各种复杂场景下视频运动目标的检测,检测准确率达到93%,优于目前的主流算法。

In order to solve moving object detection in video which is greatly influenced by background and the detec-tion effect of complex scene is not good,in this paper,a method of moving video object detection based on super-pix-el uniform saliency was proposed.Firstly,the video frame sequence is segmented into super-pixel to extract the local features and reduce the computational complexity of subsequent processing.Then,the video single frame image is de-tected significantly to obtain the image's salient target area;and then the moving object is detected by the super-pixel matching mechanism between the video frames.Finally,the synergetic discriminant factor of video inter-frame saliency propagation is introduced to improve the discrimination accuracy of moving objects.The results show the proposed algo-rithm has strong robustness and is suitable for the detection of moving video objects in complex scenes.The detection accuracy reaches 93%,which is better than the current mainstream algorithms.