基于高斯混合背景模型的运动目标检测技术研究

摘要

在运动目标检测技术中,使用传统的高斯混合背景模型所得到的检测结果并不能完美地获取运动目标的轮廓信息,而图像中像素的梯度信息,刚好就是反映了各物体的轮廓和边界,并且相对于颜色信息而言,梯度信息对于噪声并不敏感。为此,该文对传统的高斯混合背景模型进行了改进,提出基于梯度时空信息的高斯混合背景模型,证明了改进的算法确实能够取得更好的运动目标的检测结果。%In the moving object detection algorithms,Gaussian mixture background model using the obtained test results are not perfect for moving object outline information.Gradient information of pixels in the image is just a reflection of the outline of objects and convenience,and relative to the color information that is concerned,and it is not sensitive to noise.In this paper,the background of the traditional Gaussian mixture model was proposed to improve the temporal and spatial information based on the gradient of the Gaussian mixture background model.The validation gives a proof of the improved algorithm which can indeed achieve better target detection results.