一种改进的自适应混合高斯建模算法

A Modified Self-adaptive Gaussian Mixture Modeling Algorithm

车辆识别系统对于智能交通系统具有重要的意义,也是其他技术实现或判决的重要基础之一。混合高斯模型在应对背景中存在扰动的情况时具有明显的优势,但传统的混合高斯建模算法适应场景突变的能力不强,容易产生较长时间的虚影。本文在传统算法的基础上,对背景更新过程做了改进,从而可以快速地去除不再符合要求的背景模型。实验表明,在光照发生变化或摄像头轻微抖动等情况下具有良好的自适应性,配合阴影去除算法将大幅提高车辆识别的准确率。

Vehicle identification system has significant implication for intelligent transportation systems which is also one of the foundations of other related technology. The traditional Gaussian mixture model has obvious advantages in dealing with the background disturbance, but lack of ability to adapt to the scene mutation and avoid ghost for a long time. Based on the traditional method, this paper improves the updating formula to remove the undesirable background model quickly. As the experiments show, the algorithm can significantly improve the accuracy of the Vehicle Identification and has a good self - adaptability with shadow removal algorithm under the conditions of illumination changes or camera jitter.