提高目标检测算法在复杂场景下的检测鲁棒性是目前计算机视觉领域的一个重点、难点问题。传统的运动目标检测都是基于目标的单一特征, 提出一种融合颜色特征和纹理特征的背景建模方法, 并将其运用于运动目标检测。首先对基于颜色的高斯混合模型加以改进, 减少了传统高斯混合模型的计算量, 然后将高斯混合模型与LBP纹理模型用D-S证据理论进行融合。实验结果表明, 两个特征的融合有很好的互补作用, 并且能够实时、准确地检测出运动目标。

To improve the robustness of target detection under complex scene is an important and difficult challenge in the field of computer vision.The traditional moving target detection is based on the single feature of the target.In this paper,a new background modeling algorithm is proposed for object detection based on both color feature and texture feature.Firstly,the traditional color-based Gaussian mixture model is improved and the computation cost is reduced.Then the improved Gaussian mixture model is fused with the LBP texture model by D-S evidence theory.The experimental results show that the fusion of two features can be complementary.The proposed algorithm can detect the moving targets more rapidly and accurately compared with the traditional algorithm.