中文摘要：

背景减除是机器视觉中运动目标检测最常用的方法。针对复杂场景中传统单特征算法检测精度不高、多特征检测算法实时性较差的问题,提出了一种改进的联合纹理、颜色和位置特征的局部联合特征,并对局部联合特征混合高斯建模,采用多重判定进行学习和更新的目标检测算法。为更好地抵抗阴影和光照变化影响并减少计算量,改进了LBP算子,Lab局部颜色特征在处理纹理缺乏情况时,有更好的效果,而位置特征能减缓场景抖动等噪声影响。实验结果表明,该算法能准确地检测上述影响下的目标,检测效果在多种数据集上表现出更高的鲁棒性和精确性并且基本达到实时性要求。

英文摘要：

Background subtraction is the most commonly used method in machine vision for moving object detection. Aiming at the problem that the detection accuracy of traditional single feature algorithm is not high and the real-time performance of multiple feature detection algorithm is poor in complex scene,an improved object detection method based on local united feature is proposed in this paper. The method combines texture,color and location features; performs Gaussian mixture modeling for the local united feature; and adopts multiple determinations to conduct learning and update. In order to resist the influence of shadow and lighting changes as well as reduce the amount of computation,the（ local binary pattern LBP） operator is modified. The Lab local color feature performs better in dealing with the case of lacking texture. Meanwhile,the location feature can mitigate the influence of scene jitter. Experiment results show that the proposed algorithm can accurately detect the target under the influences mentioned above; the detection results exhibit higher robustness and accuracy on a variety of detection datasets; in addition,the algorithm achieves the real-time requirement basically.