【Review】用于SLAM的地点识别(Place Recognition)

1. Point Cloud(Lidar)

2021

ICRA

  1. NDT-Transformer: Large-Scale 3D Point Cloud Localisation using the Normal Distribution Transform Representation
    (Paper, Github)
  2. DiSCO: Differentiable Scan Context with Orientation
    (paper, Github)
  3. Locus: LiDAR-Based Place Recognition Using Spatiotemporal Higher-Order Pooling
    (paper, Github)
  4. Robust Place Recognition Using an Imaging Lidar
    (paper, Github)

ICCV

  1. Pyramid Point Cloud Transformer for Large-Scale Place Recognition
    (Paper, Github)

IROS

  1. SSC: Semantic Scan Context for Large-Scale Place Recognition
    (Paper, Github)
  2. A Registration-aided Domain Adaptation Network for 3D Point Cloud Based Place Recognition
    (Paper, Github)
  3. Evaluation of Long-Term LiDAR Place Recognition

CVPR

  1. A Registration-aided Domain Adaptation Network for 3D Point Cloud Based Place Recognition
    ((Paper, Github))

Others

  1. MinkLoc3D: Point Cloud Based Large-Scale Place Recognition
    (WACV, paper, Github)
  2. MinkLoc3D-SI: 3D LiDAR place recognition with sparse convolutions, spherical coordinates, and intensity
    (paper, Github)

2020

2019

ICCV

  1. LPD-Net: 3D Point Cloud Learning for Large-Scale Place Recognition and Environment Analysis
    (Paper, Github)

2018

IROS

  1. Stabilize an Unsupervised Feature Learning for LiDAR-based Place Recognition
    (Paper, Github)
  2. Scan Context: Egocentric Spatial Descriptor for Place Recognition within {3D} Point Cloud Map
    (Paper, Github)

2. Researchers

  1. Peng Yin
    Carnegie Mellon University | CMU · Robotics Institute
  2. Ji Zhang
    Carnegie Mellon University | CMU · Robotics Institute
  3. Giseop Kim
    Korea Advanced Institute of Science and Technology | KAIST · Department of Civil and Environmental Engineering
  4. Lin Li
    Zhejiang University | ZJU · Department of Control Science and Engineering

3. Visual

2021

IROS

  1. A Hierarchical Dual Model of Environment- and Place-Specific Utility for Visual Place Recognition
    (Paper, Github)
  2. Visual Place Recognition using LiDAR Intensity Information
    (Paper, Github)
  3. SymbioLCD: Ensemble-Based Loop Closure Detection using CNN-Extracted Objects and Visual Bag-of-Words
    (Paper, Github)
  4. Visual Place Recognition Using LiDAR Intensity Information
    (Paper, Github)

ICCV

  1. Conformer: Local Features Coupling Global Representations for Visual Recognition
    (Paper, Github)
  2. Attentional Pyramid Pooling of Salient Visual Residuals for Place Recognition
    (Paper, Github)

CVPR

  1. OBoW: Online Bag-of-Visual-Words Generation for Self-Supervised Learning
    (Paper, Github)
  2. Patch-NetVLAD: Multi-Scale Fusion of Locally-Global Descriptors for Place Recognition
    (Paper, Github)

ICRA

  1. Intelligent Reference Curation for Visual Place Recognition Via Bayesian Selective Fusion
    (Paper, Github)
  2. Appearance-Based Loop Closure Detection Via Bidirectional Manifold Representation Consensus
  3. SoftMP: Attentive Feature Pooling for Joint Local Feature Detection and Description for Place Recognition in Changing Environments
  4. Simultaneous Multi-Level Descriptor Learning and Semantic Segmentation for Domain-Specific Relocalization
  5. Resolving Place Recognition Inconsistencies Using Intra-Set Similarities
  6. Spherical Multi-Modal Place Recognition for Heterogeneous Sensor Systems
    (Paper, Github)
  7. Retrieval and Localization with Observation Constraints
  8. A Flexible and Efficient Loop Closure Detection Based on Motion Knowledge
  9. Semantic Reinforced Attention Learning for Visual Place Recognition
  10. STA-VPR: Spatio-Temporal Alignment for Visual Place Recognition
    (Paper, Github)
  11. Visual Place Recognition Via Local Affine Preserving Matching

2020

CVPR

  1. SuperGlue: Learning Feature Matching with Graph Neural Networks
    (Paper, Github)
  2. CORAL Colored Structural Representation for Bi-Modal Place Recognition
    (Paper, Github)

Reference:

  1. ICRA2021 SLAM方向论文汇总
  2. IROS2021 SLAM方向论文汇总
  3. CVPR 2021 SLAM 相关论文汇总
  4. ICCV2021 SLAM 点云配准 深度估计 自动驾驶等文章汇总
### 基于SLAM的车道识别技术概述 #### 背景与目标 在自动驾驶领域,同时定位与建图(Simultaneous Localization and Mapping, SLAM)是一种关键技术。它不仅能够帮助车辆感知环境并构建地图,还能实现实时定位功能。对于车道识别而言,结合SLAM算法可以通过提取道路标记、车道线等特征信息完成高精度的全局定位和导航。 一种典型的基于SLAM的车道识别方法是利用道路标记作为主要特征源[^1]。这种方法通过从相机图像中提取丰富的道路标记信息,并将其分类为不同的类别(如停止线、导向箭头等)。这些标记被进一步组合成子地图结构,从而增强环路闭合检测的能力。具体来说: - **子地图定义**:将道路标记及其周围的车道信息抽象为子地图单元。 - **随机森林分类器**:通过对道路标记进行六类划分,提升匹配过程中的鲁棒性和准确性。 - **误差校正机制**:通过补偿惯性测量单元(IMU)、轮速计等传感器中的累积漂移误差,显著提高了系统的整体性能。 实验表明,在约4.7公里的实际测试场景下,该方法达到了平均1.098米的全局定位精度。 --- #### 技术实现细节 ##### 数据采集与预处理 为了支持高效的车道识别,通常需要依赖多种传感器协同工作。例如: - 使用单目或双目摄像头捕获前方视野内的图像数据; - 结合IMU记录加速度变化趋势; - 利用车辆自带的速度信号补充动态参数。 针对上述输入流,需执行必要的滤波操作去除噪声干扰,随后送入后续模块作进一步分析。 ##### 特征提取与描述符生成 在此阶段,重点在于如何有效表征每帧画面中存在的关键要素——即各类交通标志图案及边界线条轮廓。以下是几种常见策略的应用实例: - ORB特征点改进版:相比原始版本仅关注像素强度梯度特性外,还引入了额外约束条件确保其具备尺度和平面旋转不变属性[^4]。 ```python import cv2 def extract_orb_features(image): orb = cv2.ORB_create() keypoints, descriptors = orb.detectAndCompute(image, None) return keypoints, descriptors ``` - CNN模型预测路面纹理分布概率密度函数曲线形状差异区分不同材质区域[^2]。 ##### 后端优化流程设计思路说明文档链接地址如下所示: 由于篇幅所限无法详尽展开论述,请参阅参考资料获取更多详情[^5]。 --- #### 应用案例分享 实际工程项目往往综合考虑软硬件配置情况选取最适合的技术路线。比如某些高端车型可能配备激光雷达辅助测距以弥补光学设备受天气状况影响较大这一短板;而低端产品则倾向于单纯依靠成本低廉的传统影像传感装置达成基本需求即可满足预期效果评估标准要求。 ---
评论
添加红包

请填写红包祝福语或标题

红包个数最小为10个

红包金额最低5元

当前余额3.43前往充值 >
需支付:10.00
成就一亿技术人!
领取后你会自动成为博主和红包主的粉丝 规则
hope_wisdom
发出的红包
实付
使用余额支付
点击重新获取
扫码支付
钱包余额 0

抵扣说明:

1.余额是钱包充值的虚拟货币,按照1:1的比例进行支付金额的抵扣。
2.余额无法直接购买下载,可以购买VIP、付费专栏及课程。

余额充值