论文部分内容阅读
目的前景检测是视频监控领域的研究重点之一。LOBSTER(local binary similarity segmenter)算法把Vi Be(visual background extractor)算法和LBSP(local binary similarity patterns)特征结合起来,在一般场景下取的了优良的检测性能,但是LOBSTER算法在动态背景下适应性差、检测噪声多。针对上述问题,提出一种改进的LOBSTER算法。方法在模型初始化阶段,计算各像素的LBSP特征值,并分别把像素的灰度值和LBSP特征值添加到各像素的颜色背景模型与LBSP背景模型中,增强了背景模型的描述能力;在像素分类阶段,根据背景复杂度自适应调整每个像素在颜色背景模型和LBSP背景模型中的分类阈值,降低了前景中的噪声;在模型更新阶段,根据背景复杂度自适应调整每个像素背景模型的更新策略,提高背景模型对动态背景的适应能力。结果本文算法与Vi Be算法和LOBSTER算法进行了对比实验,本文算法的前景图像比Vi Be算法和LOBSTER算法的噪声点大幅较低,本文算法的PCC指标在不同视频库中比Vi Be算法提高0.736%7.56%,比LOBSTER算法提高0.77%12.47%,FPR指标不到Vi Be算法和LOBSTER算法的1%。结论实验仿真结果表明,在动态背景的场景下,本文算法比Vi Be算法和LOBSTER算法检测到的噪声少,具有较高的准确率和鲁棒性。
The purpose of foreground detection is one of the research focuses in the field of video surveillance. LOBSTER (local binary similarity segmenter) algorithm combines the features of Vi Be (visual background extractor) and LBSP (local binary similarity patterns), and has good detection performance in general scenarios. However, the LOBSTER algorithm has poor adaptability in dynamic context , Detection of noise and more. Aiming at the above problems, an improved LOBSTER algorithm is proposed. Methods The LBSP eigenvalue of each pixel was calculated during the initial stage of the model and the pixel gray value and LBSP eigenvalue were added respectively to the color background model and the LBSP background model of each pixel to enhance the description ability of the background model. In the classification stage, the classification threshold of each pixel in the color background model and the LBSP background model is adaptively adjusted according to the background complexity to reduce the noise in the foreground. In the model updating stage, each pixel background model is adaptively adjusted according to the background complexity Update strategy to improve the adaptability of the background model to the dynamic context. Results Compared with Vi Be algorithm and LOBSTER algorithm, the proposed algorithm has a significantly lower foreground image noise ratio than Vi Be algorithm and LOBSTER algorithm. The PCC index of this algorithm is 0.736 higher than Vi Be algorithm in different video libraries % 7.56%, which is 0.77% and 12.47% higher than LOBSTER algorithm. FPR index is less than 1% of Vi Be algorithm and LOBSTER algorithm. Conclusion The experimental simulation results show that the proposed algorithm has less noise than Vi Be algorithm and LOBSTER algorithm in the dynamic background scenario, with high accuracy and robustness.