Abstract:Vehicle-to-Vehicle (V2V) cooperative perception has become increasingly popular in the field of autonomous driving, effectively overcoming the inherent limitations of single-vehicle perception systems, such as limited range and susceptibility to occlusions. In a V2V system, vehicles in close proximity can share perception data. To fuse this data, which is collected from different viewpoints by each vehicle, accurate pose information (including position and heading direction) is essential to transform the received data to the receiving vehicle's viewpoint. However, pose errors, often caused by measurement noise or sensor failures, can lead to severe misalignment during data fusion, resulting in incorrect object detections and potentially hazardous decisions in autonomous driving systems. To address this challenge, we present BB-Align, a lightweight pose recovery framework that utilizes Lidar Bird's-eye View (BV) images and object bounding Boxes for relative pose estimation. Designed as a plug-and-play solution, the proposed method requires no additional model training, enabling effortless integration into existing V2V systems. Our approach uses Lidar-derived BV images with a Log-Gabor filter-based feature map for effective image matching despite image sparsity. To reduce errors from self-motion distortion, we also integrate object bounding boxes for finer alignment. The proposed method is rigorously evaluated on the V2V4Real dataset-currently the only real-world V2V dataset. Our approach demonstrates high pose estimation accuracy, outperforming an existing graph-matching method. It achieves translation and rotation errors of less than 1 m and 1., respectively, in 80% of cases within a 70 m range between vehicles. Furthermore, by integrating the proposed framework into cooperative object detection models under serious pose error, the result shows up to a 2x increase in Average Precision (AP) compared to those without pose recovery, with more pronounced improvements in the short range.

Automatic and Accurate Vision-Based Measurement of Camber and Toe-In Alignment of Vehicle Wheel

BB-Align: A Lightweight Pose Recovery Framework for Vehicle-to-Vehicle Cooperative Perception

Deterministic Optimality for Robust Vehicle Localization Using Visual Measurements

Large-Scale All-Wheel Alignment Globally Registered by Laser Line Family and Verified by Global Benchmark

Vision measurement system for position-dependent geometric error calibration of five-axis machine tools

Methods of Automatic Calibration for Four-Wheel Alignment Based on Superimposing Angles

Optimization Scheme of Measuring Accuracy of Bearing Seat Size Based on Machine Vision

Research On The Precise Calculation And Control Strategy Of Front Wheel Toe-In Of Vehicles

Inertial Measurement System For Track Alignment Inspection Based On Machine Vision

Machine Vision-Based Method for Reconstructing the Vehicle Coordinate System in End-of-Line ADAS Calibration

Wheel Alignment Based On Computer Vision Without Pushing Car Technique Research

Camera-Mirror Binocular Vision-Based Method for Evaluating the Performance of Industrial Robots

Automated Measurement and Hybrid Adaptive Identification Method for Kinematic Calibration of Hybrid Machine Tools

Mounting Misalignment and Time Offset Self- Calibration Online Optimization Method for Vehicular Visual-Inertial-Wheel Odometer System

High Precision Vehicle Localization Based on Tightly-coupled Visual Odometry and Vector HD Map

Modeling and Statistical Analysis of Vision Distance Measurement Uncertainty for Multi-Automotive Sensor Target Tracking

Vehicle Lateral Dynamics Modelling by Subspace Identification Methods and Tyre Cornering Stiffness Estimation

Factors affecting the accuracy of a computer vision-based vehicle weight measurement system

Analysis of the Precise Calculation of Toe-in of Front Wheel Based on a 1026 Model Truck

Research on online calibration of lidar and camera for intelligent connected vehicles based on depth-edge matching

Measurement of articulation angle by image template matching