基于高斯混合模型的改进GVF-Snake运动目标检测算法

Improved GVF-Snake Moving Object Detection Algorithm Based on Gauss Mixed Model

针对GVF-Snake算法没有使用图像中统计信息的缺陷,提出了改进算法,增强算法在复杂背景下的检测效果。改进算法通过引入混合高斯模型,对复杂背景进行建模。通过对目标图像与背景模型进行差分,从而构造得到新的外部能量项,计算得到GVF场,用以检测运动目标的边缘。此外,通过对第一帧图像和高斯混合模型背景做差分,再经过形态学处理即可得到较好的初始轮廓,增强了算法的自适应能力。应用本文提出的改进算法对运动目标做检测实验,结果表明,本文提出的算法对于复杂背景下的变形目标具有较好的检测效果。

Considering the defect of GVF-Snake algorithm without using image statistics,the improved algorithm is proposed to enhance the detection effect of algorithm under complex background. The improved algorithm realize the model of complex background by using Gauss mixed model. Through the difference of object image and background model,the new external energy term is gained and GVF field is computed,which are used to detect the edge of moving object. In addition,the first frame image can obtain the better initial contour by using the difference with the Gauss mixed model background and morphological processing,increasing the adaptive ability. Experimental results show that the proposed algorithm obtains better detect effect for deformation object under complex background.