基于改进Unet的起重机攀爬机器人车道识别

Lane recognition of crane climbing robot based on improved Unet

起重机攀爬机器人是一种代替人力检测起重机的机械设备。为解决起重机攀爬机器人自动寻路的车道识别问题,文中设计了一种针对起重机金属结构特点优化,将MobilenetV2作为特征提取部分与Unet网络相结合的M2-Unet卷积神经网络。用攀爬机器人在门式起重机上采集1979张图像数据,由专业标注软件Labelme制作成数据集进行训练与测试,并使用其他2种主流的语义分割网络在相同的数据集上作对比实验。实验结果表明,相较于其他2种图像分割网络,改进的M2-Unet卷积神经网络的分割准确率最高;M2-Unet网络对测试集479张图片的平均识别准确率在96%以上,平均运行时间远小于0.5 s,能同时满足起重机攀爬机器人车道识别任务的实时性和精度要求。

Crane climbing robot is a kind of mechanical equipment to replace cranes for manual inspections.To solve the problem of lane recognition in automatic path finding of crane climbing robot,this paper proposes an optimization scheme as per characteristics of crane metal structure.Taking Mobilenet V2 as the feature extraction part,and combining M2-Unet convolutional neural network with Unet network,a climbing robot was used to collect 1979 images from the gantry crane,and the data set was made by the professional labeling software Labelme,which was trained and tested.Two other mainstream semantic segmentation networks were used to conduct comparative experiments on the same data set.Experimental results show that compared with other two image segmentation networks,the improved M2-Unet convolution neural network has the highest segmentation accuracy.M2-Unet network has an average recognition accuracy of over 96%for 479 images in the test set,and its average running time is much shorter than 0.5 s,which can meet the requirements of real-time and accuracy of lane recognition of crane climbing robot.