基于改进的背景差分的运动目标实时检测算法

针对传统混合高斯建模算法计算量过大与目标轮廓清晰度小的问题，提出了一种新的运动目标实时检测算法。该算法引入三帧差分的方法，提高了检测目标轮廓的清晰度；通过HSI混合高斯建模前进行分块处理有效减小了计算量，因此算法的实时性有了明显的改善；并利用逻辑运算融合三帧差分与HSI混合高斯模型进行高效的背景提取；最后运用数学形态学方法进一步优化检测结果。实验结果表明，相比混合高斯模型经典算法，该算法能更快速、更准确地检测出智能监控视频序列中的运动目标，并且目标轮廓清晰度也有明显的改善。

Moving objects real time detection algorithm based on the improved background subtraction

The traditional mixture Gaussian models require much computation and have little clarity of objects’ contours. Therefore, a new moving objects real time detection algorithm is proposed. Firstly, the three frame differencing method is introduced in the algorithm in order to improve the contours’ clarity of detection objects. Secondly, the computation is reduced by block processing in the Gaussian mixture models on HSI, so the real time performance of the algorithm is improved. Thirdly, the three frame differencing method and the adaptive Gaussian mixture model on HSI are merged by logic computation so as to extract background efficiently. Finally, the detection result is optimized further by the mathematical morphology. The experimental results show the new algorithm can detect the moving objects in the surveillance video sequences faster and more accurately than the classic mixture Gaussian models and improve the clarity of objects’ contours.