一种改进的基于混合高斯模型的运动目标检测算法

运动目标检测是实现目标跟踪和行为分析等任务的基础。在运动目标检测中,消除背景与噪声的干扰,从而将运动目标从图像中分离出来一直是研究的重点。混合高斯模型法被广泛地应用于运动目标检测,对存在小幅度运动的背景有较好的抗干扰能力,并且能提取出较完整的运动目标,但是同时存在噪声干扰,且对阴影抑制效果较差。针对传统混合高斯模型法的不足,提出一种改进的基于混合高斯模型的运动目标检测算法,利用帧差法对光照突变适应性较好和算法简单的特点,将传统混合高斯模型法与和四帧差法结合。实验结果表明,该方法能够有效地消除复杂环境中的噪声,并对阴影有一定的抑制作用,提高了运动目标检测的准确性和完整性。

Improved Moving Target Detection Algorithm Based on Gaussian Mixture Model

Moving object detection is the basis for tracking and behavior analysis tasks.In moving target detection,eliminating the interference of background and noise and separating moving targets out from the image have been the focus of the study.Gaussian mixture model method is widely used in object detection, has a better anti-interference ability for existence of small amplitude motion of background and can extract more complete moving target,but at the same time noise exists and shadow suppression is less effective.An improved algorithm for moving target detection based on Gaussian mixture model was proposed to make up the deficiencies of original Gaussian mixture model method,original Gaussian mixture model and four-frame differencing were combined by taking advantage of the better adaptability to light change and simple algorithm features of frame difference method.The experimental results indicate that the proposed method can eliminate noise and shadow effectively in complex environment and improve the accuracy and integrity of moving target detection.