基于随机选择和竞争合作更新策略的变化检测

针对复杂场景中背景更新这一难题，提出一种背景更新适应性较强的变化检测算法。利用建模样本的无记忆性进行随机更新得到初步更新后的背景模型，采用模型像素间竞争的方式选取邻域更新位置，通过当前帧图像像素间合作的方式得到加权更新值。仿真实验结果表明，所提更新策略能够在有效处理复杂场景的同时保证检测结果的准确率。五类复杂背景图像序列的仿真结果也验证了该算法具有综合性能的优势。

Change detection based on randomly selective and co-competitive update strategy

This paper presents a model update method for change detection which can adapt strongly in complex background. Random update is used for preliminary model update, since samples are memoryless. It competitively looks for the substituted pixels in the neighborhood of adapted background model. The assignment of weighted values behaves as a cooperative way between current frame pixels. Efficiency figures show that the proposed algorithm is feasible and accurate for complex background. The results of five kinds of image sequences with complex background prove excellent methods in terms of comprehensive detection rate.