摘要
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。为了实现在多种背景扰动以及阴影同时存在的复杂场景下,对运动目标的准确、鲁棒提取,论文提出了一种融合纹理特征和背景建模的自适应目标检测混合算法。首先,为了对阴影进行有效处理,论文提出融合纹理特征的背景建模法;同时,在背景建模的基础上,引入亮度信息预处理程序;最后,论文在对复杂场景下(包括室内、室外)的背景扰动进行分析归类的基础上,将帧间差分法和背景建模法有机结合,有效提高算法对复杂场景的适应性。实验结果表明,复杂场景下,该算法对大多数背景扰动都具有一定的鲁棒性,能够实时、准确地检测出运动目标。
Improving the robustness of targets-detecting algorithm under complicated scenes is an important and difficult research prob- lem in the field of computer vision. In order to achieve accurate and robust detecting result under complex scenes, with all kinds of back- ground disturbance and shadow, an adaptive targets-detecting algorithm based on LBP and background modeling method（BMM） is proposed in this paper. Firstly, BMM combined with LBP, which is less influenced by shadow than traditional Color-based BMM, is presented. Sec- ondly, light information pretreatment is proposed for situation of sudden brightness changes. Finally, an adaptive detecting mechanism is proposed. Experimental results show that, the proposed algorithm has robustness for most background disturbances, effectively improved a- daptability, real-time performance and accuracy of detecting effect under complex scenes.
出处
《计算机与数字工程》
2013年第7期1081-1084,共4页
Computer & Digital Engineering
基金
宝鸡文理学院科学研究项目(编号:YK1035)资助
关键词
背景扰动
LBP纹理
背景建模
自适应目标检测
background disturbance
LBP
background modeling
adaptive target detection
