改进自适应混合高斯模型和帧间差分的运动目标检测

Improved Adaptive Gaussian Mixture Model and Inter-frame Difference for Moving Target Detection

针对传统混合高斯模型进行运动目标检测易受环境噪声或光照变化干扰,检测结果存在空洞、边缘缺失等问题,提出一种改进自适应混合高斯模型和帧间差分的运动目标检测算法。在混合高斯模型建立之初,采用较大的学习率快速消除背景干扰信息,当模型趋于稳定后,根据目标运动状态不断调整学习率,实现自适应修正背景模型。同时引入光照变化因子调整模型更新率,有效克服光照变化的影响。为提高目标边缘连续性,采用基于图像相似度的四帧差分法并结合边缘检测算法快速提取目标轮廓信息填补目标边缘。通过形态学处理消除目标空洞及残余噪点,获得完整的运动目标。结果表明:该方法目标检测准确率达到95.2%,在保证实时性的同时对复杂场景具有较好的适应能力,鲁棒性强。

In view of the problems that the traditional Gaussian mixture model is easy to be disturbed by environmental noise or illumination change,and the detection results have holes and missing edges,a moving target detection algorithm based on adaptive Gaussian mixture model and inter-frame difference was proposed.At the beginning of the establishment of Gaussian mixture model,a large learning rate was used to quickly eliminate the background interference information.When the model tended to be stable,the learning rate was continuously adjusted according to the target motion state to realize adaptive correction of the background model.At the same time,a light change factor was introduced to adjust the model update rate and effectively overcome the influence of light change.In order to improve the continuity of the target edge,the four-frame difference method based on image similarity and edge detection algorithm were used to extract the contour information of the target and fill the target edge.The target cavity and residual noise were eliminated by morphological processing to obtain a complete moving target.The results show that the target detection accuracy of this method reaches 95.2%,and it has good adaptability and robustness to complex scenes while ensuring real-time performance.