改进MOG-LRMF的铁轨动态异物检测

Real-time Detection of Rail Dynamic Foreign Object Intrusion Based on Improved MOG-LRMF

针对复杂铁路环境下动态入侵异物检测精度低和抗扰能力差等问题,提出一种基于改进MOG-LRMF算法的铁路轨道异物入侵实时检测方法.引入仿射变换,对视频序列可能出现的抖动进行预校正处理;分析MOG-LRMF模型特点,利用MOG模型对视频帧中的背景进行建模,用前一帧背景中学习到的知识对当前帧背景进行预测,优化MOG-LRMF参数求解模型;利用EM算法对改进MOG-LRMF模型进行参数求解,实现背景在线实时更新.实验结果表明,改进的MOG-LRMF算法在光照充足、光线较弱、相机存在抖动、背景复杂及存在多个目标情形下都能提高目标检测精度,具有较好的抗干扰性、鲁棒性和快速性.

To address the issues of low detection accuracy and poor anti-interference ability for the dynamic intrusion of foreign objects in complex rail environments, a real-time detection method for foreign object intrusion in railway track based on improved MOG-LRMF algorithm is proposed in this paper. Firstly, the affine transformation is used to pre-correct video sequences. Then, the background of the frame in a video sequence is predicted with the background knowledge learned in the previous frame to improve the MOG-LRMF model by analyzing the characteristics of the MOG-LRMF model. Finally, the EM algorithm is used to solve the parameters of the MOG-LRMF model, and it can realize the online real-time update of the background. The experiment results show that the improved MOG-LRMF algorithm can greatly enhance the target detection accuracy under sufficient illumination, weak light, camera jitter, complex environment, and multiple targets. Moreover, the improved MOGLRMF algorithm has better anti-interference, robustness, and rapidity.