非理想条件下融合感知仿真数据集
收藏资源简介：
本数据集基于 CARLA 仿真平台构建，旨在对自动驾驶场景下视觉 SLAM 定位技术在非理想环境中的性能指标进行系统性数据支撑与分析。数据集主要涵盖视觉 SLAM 定位精度的仿真与测试数据，包括不同场景与不同环境条件下的定位轨 迹、位姿估计误差以及系统运行性能等内容，同时对数据采集方法、数据处理流程及数据质量控制与评估进行了规范化设计。数据采集过程中，基于 CARLA 平台分别构建了雨天、雾天、暗光及晴天等多种典型非理想与理想天气场景，对单车视 觉 SLAM 定位系统开展仿真测试，重点评估其在复杂成像条件下的定位精度与稳定性。数据仿真传感器参数严格对齐高精度车载传感器的技术规格，确保数据精度与可重复性。针对采集数据中可能引入的仿真噪声与异常值，进一步开展了必要的 数据清洗与噪声消除处理。该数据集为评估基于视觉 SLAM 的自动驾驶车辆在非理想天气条件下的高精度定位能力与鲁棒性提供了可靠的数据基础，并可为后续相关算法研究与系统优化提供参考与指导
This dataset is constructed based on the CARLA simulation platform, aiming to provide systematic data support and analysis for the performance metrics of visual SLAM positioning technology in autonomous driving scenarios under non-ideal environments. The dataset mainly covers simulation and test data of visual SLAM positioning accuracy, including positioning trajectories, pose estimation errors, system operating performance and other contents under different scenarios and environmental conditions. Meanwhile, it has standardized designs for data collection methods, data processing workflows, as well as data quality control and evaluation. During the data collection process, multiple typical ideal and non-ideal weather scenarios such as rainy, foggy, low-light and sunny conditions were built based on the CARLA platform, and simulation tests were conducted on single-vehicle visual SLAM positioning systems, focusing on evaluating their positioning accuracy and stability under complex imaging conditions. The parameters of the simulated sensors are strictly aligned with the technical specifications of high-precision vehicle-mounted sensors, ensuring data accuracy and repeatability. For the simulation noise and outliers that may be introduced in the collected data, necessary data cleaning and noise elimination processing were further carried out. This dataset provides a reliable data basis for evaluating the high-precision positioning capability and robustness of autonomous vehicles based on visual SLAM under non-ideal weather conditions, and can also provide references and guidance for subsequent relevant algorithm research and system optimization.
