卷积神经网络在危化品仓库智能巡检车中的应用研究

刘学君¹，李齐飞¹，沙芸¹，甘建旺¹，隗立昂¹

1. 北京石油化工学院信息工程学院，北京市，102617

摘要：尽管巡检车的研究成果突飞猛进，但目前取得的成就对硬件要求极为苛刻。因此，在危化品仓库智能巡检车只包含摄像头及必要硬件的条件下，结合视觉与深度学习，将卷积神经网络应用在危化品仓库巡检车道路样本处理上，经过调整超参数及对比一系列实验结果，选取适用于智能巡检车的参数及网络架构，用于获取智能巡检车行驶速度和车轮应转角度。

关键词：卷积神经网络；危化品仓库；智能巡检车；深度学习；Lenet-5；超参数

Researchonapplicationofconvolutionalneuralnetwork inintelligentinspectionvehicleofchemicalwarehouse

Liu Xuejun¹，Li Qifei¹，Sha Yun¹，Gan Jianwang¹，Wei Li-ang¹

1. School of Information Engineering, Beijing Institute of Petrochemical Technology, Beijing City, 102617

Abstract：Despite the rapid progress of the research on the inspection vehicle, the current achievements are extremely demanding on the hardware. Therefore, under the condition that the hazardous chemicals warehouse intelligent inspection vehicle only contains the camera and the necessary hardware, combined with visual and deep learning, the convolutional neural network is applied to the road sample processing of the hazardous chemicals warehouse inspection vehicle, and the super parameters are adjusted. Comparing a series of experimental results, this puper selects, the parameters and network architecture suitable for the smart inspection vehicle, which are used to obtain the speed of the intelligent inspection vehicle and the angle of rotation of the wheel.

Keywords: Convolutional neural network; Hazardous chemical warehouses; Intelligent patrol inspection vehicle;Deep learning;Lenet-5;Super parameter