城市轨道交通小型障碍物检测

Detection of small obstacles in urban rail transit

障碍物是影响列车行车安全的重要因素,小型障碍物由于体积小,在检测中容易被遗漏。针对上述问题,提出了一种基于卷积神经网络的自动化检测算法。算法首先通过数据增强策略平衡样本数量,然后使用卷积网络进行特征融合,很好地结合了小型障碍物的位置信息和语义信息。实验结果表明,算法对列车前方的小型障碍物有良好的检测效果,能较好地实现小型障碍物的探测。

Obstacles are an important factor affecting train operation safety.Due to little size,small obstacles can be ignored easily.To solve above problems,an automatic detection algorithm had been proposed based on convolutional neural network.Data augmentation was used in this algorithm to balance the sample size and then convolutional neural network is used to feature fusion,combining the location information and semantic information of small obstacles.The experimental results show that the algorithm mentioned has good detection effect on small obstacles.