基于熵和相关接近度的混合高斯目标检测算法

发布时间：2018-09-08 10:35

【摘要】：针对固定模型个数的混合高斯模型的背景建模速度慢和运动目标的拖影问题,提出了一种基于Tsallis熵和相关接近度的改进混合高斯算法。该算法利用Tsallis熵对高斯模型自适应地选择模型个数,加速背景建模;对于模型匹配判断条件,不能很好地体现相邻像素点的空间相关性的情况,提出了相关接近度作为模型更新的限定条件,以去除拖影。实验结果表明,改进的算法在实时性、检测正确率方面都有较好的改进。
[Abstract]:Aiming at the problem of slow modeling speed and drag and shadow of moving targets in the mixed Gao Si model with fixed number of models, an improved mixed Gao Si algorithm based on Tsallis entropy and correlation approach is proposed. The algorithm adaptively selects the number of models to Gao Si model by using Tsallis entropy, and accelerates the background modeling. For the model matching judgment condition, it can not well reflect the spatial correlation of adjacent pixels. The correlation proximity is used as the qualification condition of model updating to remove the drag shadow. The experimental results show that the improved algorithm has better performance in real time and detection accuracy.
【作者单位】：兰州理工大学计算机与通信学院;
【基金】：国家自然科学基金项目(61263019)资助
【分类号】：TP391.41

【相似文献】