摘要
为了使运动目标检测算法能够适应场景复杂的情况,可采用高斯混合模型进行建模的方法,本文在此基础上提出了一种新的快速检测算法。该算法基于像素时域和空域相关性,利用这一性质可以减少高斯混合模型的个数,从而提高算法运行速度,经验证发现算法的精确度没有大的降低;同时给出了一种卡尔曼预测的更新算法,这既保证了稳定性又能快速建立背景模型。实验结果表明,该算法在不降低原来算法精确度的基础上,有效地提高了算法的速度。
In order to adapt the algorithm of moving object detection to complex scene, a method of Gaussian mixture model is presented, and in light of the above a new algorithm of fast detection is proposed in this paper. This algorithm is based on correlation in temporal and spacial domains, which reduce the number of models and improve running speed, and fortunately does not lower accuracy too much; at the same time, updating method based on Kalman predicting is proposed,and the method increases the speed of background modeling and keeps stability. Additionally, experiment results show that the proposed algorithm not only guarantees accuracy but also improves detecting speed effectively.
出处
《电子测量技术》
2007年第6期33-35,79,共4页
Electronic Measurement Technology
关键词
高斯混合模型
卡尔曼预测
运动目标检测
背景减
Gaussian mixture model
Kalman predicting
moving object detection
background subtraction
