基于粒子滤波与LSTM网络对未标记AGV的追踪方法

Unmark AGV tracking method based on particle filtering and LSTM network

在自动导引车(AGV)的全局视觉导航系统中,为了实现精确的目标追踪,提出了一种针对未标记AGV的追踪方法。首先通过帧差法对运动目标进行检测,利用粒子滤波对运动目标进行跟踪;其次建立追踪目标的运动模型来学习目标的运动特征,从而预测目标丢失后的行进位置;最后利用支持向量机(SVM)模型进行预测区域内AGV的再识别以及对长短期记忆(LSTM)网络的预测结果进行校正,达到目标持续追踪。实验表明:此方法可以较好地解决由于目标遮挡、目标之间高度相似性而导致追踪失败的情况。

In order to achieve accurate target tracking in global visual navigation system of automatic guided vehicle(AGV),a tracking method for unmarked AGV is proposed.Firstly,the moving target is detected by the frame difference method,and the moving target is tracked by particle filtering.Secondly,motion model for tracking target is established to learn the motion characteristics of target,so as to predict the travel position after the target is lost.Finally,the support vector machine(SVM)model is used to re-identify the AGV in the prediction region and correct the prediction result of the long short time memory(LSTM)network to achieve the target duration tracking.Experiments show that this method can better solve the problem of tracking failure due to the high similarity between target occlusion and target.