{
  "paper_id": "54c0078f04beaa3a9211298aedd681e4",
  "paper_title": "High-Precision BEV-Based Road Recognition Method for Warehouse AMR Based on IndoorPathNet and Transfer Learning",
  "paper_authors": [
    "Tianwei Zhang",
    "Ci He",
    "Shiwen Li",
    "Rong Lai",
    "Zili Wang",
    "Lemiao Qiu",
    "Shuyou Zhang"
  ],
  "publication_date": "2024-05-27",
  "publication_venue": "Applied Sciences",
  "abstract": "The rapid development and application of AMRs is important for Industry 4.0 and smart logistics. For large-scale dynamic flat warehouses, vision-based road recognition amidst complex obstacles is paramount for improving navigation efficiency and flexibility, while avoiding frequent manual settings. However, current mainstream road recognition methods face significant challenges of unsatisfactory accuracy and efficiency, as well as the lack of a large-scale high-quality dataset. To address this, this paper introduces IndoorPathNet, a transfer-learning-based Bird's Eye View (BEV) indoor path segmentation network that furnishes directional guidance to AMRs through real-time segmented indoor pathway maps. IndoorPathNet employs a lightweight U-shaped architecture integrated with spatial self-attention mechanisms to augment the speed and accuracy of indoor pathway segmentation. Moreover, it surmounts the challenge of training posed by the scarcity of publicly available semantic datasets for warehouses through the strategic employment of transfer learning. Comparative experiments conducted between IndoorPathNet and four other lightweight models on the Urban Aerial Vehicle Image Dataset (UAVID) yielded a maximum Intersection Over Union (IOU) of 82.2%. On the Warehouse Indoor Path Dataset, the maximum IOU attained was 98.4% while achieving a processing speed of 9.81 frames per second (FPS) with a 1024 x 1024 input on a single 3060 GPU.\n",
  "keywords": [
    "road recognition",
    "semantic segmentation",
    "transfer learning"
  ],
  "paper_institutions": [
    "School of Mechanical Engineering, Yangzhou University",
    "Canny Elevator",
    "Department of Mechanical Engineering, Zhejiang University"
  ],
  "output_language": "zh",
  "provider": "doubao",
  "model": "deepseek-v4-flash",
  "prompt_template": "simple",
  "extracted_at": "2026-08-18T10:30:47.884512Z",
  "template_tag": "simple",
  "summary": "### 摘要\n\n本文针对大型动态平面仓库中AMR的视觉道路识别问题，指出现有主流方法在精度、效率以及高质量数据集缺乏方面的不足，并基于语义分割、U-Net和迁移学习等研究背景，提出IndoorPathNet：一种基于迁移学习的BEV室内路径分割网络。该网络采用轻量级U形编码器-解码器结构，使用膨胀卷积扩大感受野，并引入空间自注意力机制来整合上下文语义信息；其核心机制为通过S_ji=1/(1+exp(B_i·C_j))计算空间注意力图，再以E_j=ReLU(α∑s_ji D_i)+A_j加权聚合全局特征。为缓解仓库语义数据集稀缺，模型先在UAVID上预训练，再在自建WIPD上冻结编码器浅层并微调，同时采用边缘检测辅助任务和随机裁剪拼接数据增强。实验在UAVID和WIPD上进行，使用Accuracy、Precision、Recall、IoU等指标；与FCN-8s、FCN-32s、SegNet、U-Net相比，IndoorPathNet在UAVID上取得最高IoU 82.2%，在WIPD上最高IoU达98.4%，单张3060 GPU上1024×1024输入下处理速度约9.81 FPS（平均0.126秒/张），满足低速AMR实时全局引导需求。未来工作包括自然光照条件、多摄像头视角融合、物理尺度恢复以及与多模态SLAM结合等方向；实际应用中需考虑BEV相机、ROS、WiFi传输、0.5秒一致性校验等约束，并可通过V2X和直方图分析等扩展。\n\n### 关键要点\n\n- IndoorPathNet是为大型平面仓库AMR设计的基于BEV图像的端到端室内路径分割网络，通过实时分割室内通道图提供方向引导。\n- 网络采用U形编码器-解码器结构，编码器使用膨胀因子为2和5的膨胀卷积替代标准3×3卷积，以扩大感受野并保留关键空间信息。\n- 空间自注意力机制用S_ji=1/(1+exp(B_i·C_j))生成注意力图，并通过E_j=ReLU(α∑s_ji D_i)+A_j实现全局上下文特征聚合。\n- 为应对仓库语义数据集稀缺，采用迁移学习策略：先在UAVID上预训练，再冻结编码器前三个浅层权重，在WIPD上微调。\n- UAVID包含30个视频序列、300张高分辨率无人机城市图像，原始8类标签被合并为道路/背景二分类，并用Canny算法进行边缘检测辅助处理。\n- WIPD是使用Midjourney生成50张BEV仓库图像、用LabelMe人工标注的自建数据集，包含货物和起重设备等障碍物。\n- 数据增强采用基于Beta分布随机中心点的四图裁剪拼接方法，显著扩大训练集并提升模型对多样特征的识别能力。\n- 训练使用Adam优化器，编码器和解码器初始学习率分别为1×10^-4和1×10^-5，采用混合精度训练、L2正则化、批大小1和最多200轮迭代。\n- 损失函数结合Dice loss和交叉熵损失，以应对室内图像中像素不平衡问题。\n- 在UAVID全测试集上，IndoorPathNet的Precision、Recall和IoU分别为0.893、0.875和0.822，较FCN-8s、FCN-32s、SegNet和U-Net均有提升。\n- 在WIPD上，IndoorPathNet达到最高IoU 98.4%，处理速度约9.81 FPS（平均0.126秒/张），满足低速AMR的实时全局引导需求。\n- 实际应用流程包括BEV相机标定、WiFi传输、关键帧提取、模型分割、轮廓一致性和短时一致性校验，每0.5秒向AMR发送一次全局方向引导；当前方法仍受自然光、单视角覆盖和缺乏物理尺度等约束。"
}