一种基于随机码本的运动目标检测算法

针对传统码本算法模型建立时间过长、更新效果 差等问题,将视觉背景提取(ViBe)算法中模型建 立和更新的 思想用于改进码本算法,提出了一种基于随机码本(RCB)的运动目标检测算法。为减少计算 量,提出了一种基于 YUV空间的码本模型；为减少背景建模时间、提高模型的洁净度 ,提出了随机选取第1帧图像空间领域内 像素点的码本训练方法；为提高背景模型对复杂场景的适应能力,提出了基于随机策略的码 本更新方式。 与典型算法进行了两组实验。结果表明,本文算法兼有二者的优点,不仅能迅速适应场 景的转换,而 且在检测精度、动态适应能力和实时性等方面都有较大提高。

Moving objects detection algorithm based on random codebook

Aiming at shortening the time of building model and improving the effect of updating model in traditional codebook algorithm,with the thoughts of model establishment and updating from visual background extractor (ViBe),a new moving object detection algorithm called random codebook is put forward in this paper.Based on the traditional codebook algorithm,a new codebook model on YUV color space is raised to reduce the computational complexity,and a new codebook training method is proposed based on pi xels selected randomly in the field of space domain from the first image to shorten the model-building time and impro ve the cleanliness of it.Besides, in order to improve the adaptability of the background model to complex scenes, a random updating strategy for the codebook is also used in the new algorithm.Finally,two comparative experim ents are done with typical algorithms.And the experimental results show that the improved algorithm could not only adapt quickly to the changes of scenes but also greatly improve the det e ction accuracy,dynamic adaptability and real-time performance.