【摘要】 针对生物运动目标检测问题,为解决生物目标在运动检测中存在着非线性运动、检测速度慢以及目标易变形等问题,提出了一种基于时空熵与标签技术相结合的运动目标检测算法。方法结合标签技术将熵的相关原理运用于生物运动目标检测领域。首先对每个象素点建立一个基于时间和空间的滑动窗,按着运用标签技术对滑动窗分配标签,并求得点的直方图进而计算出点的时空熵,可利用熵的大小来判断点是否属于运动目标点,进行对生物运动目标的实时仿真。实验结果表明运算复杂度降低、参数鲁棒性较强、抗噪声能力较好,证明是一种可靠和有效的生物运动目标检测方法。更多还原

【Abstract】 In order to detect the biological moving object effectively,the paper proposed a motion detection method based on the spatial-temporal entropy and label technique.The method used the concept of entropy which came from information theory.First,a spatial-temporal sliding window was built for each pixel,and the pixels belonging to the sliding window were assigned labels according to the label technique.The label distribution in the sliding window was used to construct the entropy image in which a pixel value with constant entropy was considered as part of a moving object.Experimental results show the method is robust,has lower computational complexity and is an effective method of biological moving target detection.更多还原