铁路轨道异物完整性检测与跟踪算法研究

针对野外复杂环境下轨道异物检测不完整问题，提出基于小波变换的像素过滤思想改进GMM，构建背景模型；为解决异物目标实施机动(转弯、加速或突然出现)时跟踪实时性差和准确率低的问题，分析Kalman滤波线性化误差，搭建BP神经网络修正 IMM的跟踪模型，实现轨道异物跟踪预测，并推导出非线性Kalman滤波关系.实验表明，改进GMM在正常天气下平均前景误检率降低了24.94个百分点，针对复杂恶劣天气平均前景误检率降低了33.76个百分点；建立BP神经网络-IMM-Kalman滤波模型不仅可以快速准确地对场景中的机动目标进行跟踪，而且比Kalman滤波和IMM更加平稳，误差更小.

Detection and Tracking Algorithm of Foreign Integrity in Railway Tracks

Aiming at the problem of incomplete detection of orbital objects in complex field environment, this paper proposes a pixel filtering idea based on wavelet transform to improve GMM and construct a background model. The maneuvering (turning, accelerating or sudden appearance) is implemented to solve the foreign object target. When tracking the problem of low real-time performance and low accuracy, analyzing the linearization error of Kalman filter, BP neural network is used to modify the tracking model of IMM to realize the tracking and prediction, and the nonlinear Kalman filtering relationship is derived. The experimental shows that the average foreground false detection rate of the improved GMM is reduced by 24.94 percentage points under normal weather conditions, and the average prospect false detection rate for complex bad weather is reduced by 33.76 percentage points. The BP neural network-IMM-Kalman filtering model is not only established. It can track the maneuvering target in the scene quickly and accurately, and it is more stable and less error than the Kalman filtering algorithm and IMM.