图像及点云转码测试数据集

本数据集采集于浙江省某典型城市复杂路口，涵盖了1小时的1080p高清脱敏视频数据（MPEG格式）以及7组连续的非结构化激光雷达点云序列（BIN格式）。数据采集采用了工业级高频感知设备，视频帧率为25fps，点云帧率为10fps（100ms周期），并完整保留了三维空间坐标（x, y, z）与反射强度信息。针对边缘计算场景，本研究详细记录了基于硬件加速环境下的转码测试流程，采用前沿压缩算法进行基准测试，验证了数据在硬件加速架构下的吞吐性能。数据集经过了严格的时间戳对齐、空间有效性验证及隐私脱敏处理。本数据集的发布填补了路侧边缘计算领域缺乏专用转码基准数据的空白，为研究低延迟压缩协议、优化MEC节点算力配置以及构建大规模交通数字孪生系统提供了关键的数据支撑。

This dataset was collected at a typical complex urban intersection in Zhejiang Province, encompassing 1 hour of 1080p high-definition anonymized video data (in MPEG format) and 7 consecutive unstructured LiDAR point cloud sequences (in BIN format). The data was collected using industrial-grade high-frequency sensing equipment, with a video frame rate of 25 fps and a point cloud frame rate of 10 fps (100 ms cycle), and fully preserved the 3D spatial coordinates (x, y, z) and reflection intensity information. For edge computing scenarios, this study thoroughly documented the transcoding test procedures under a hardware-accelerated environment, conducted benchmark tests using cutting-edge compression algorithms, and validated the throughput performance of the data on the hardware-accelerated architecture. The dataset has undergone strict timestamp alignment, spatial validity verification, and privacy anonymization processing. The release of this dataset fills the critical gap in the lack of dedicated transcoding benchmark data for the roadside edge computing domain, providing pivotal data support for research on low-latency compression protocols, optimization of MEC node computing resource allocation, and construction of large-scale traffic digital twin systems.

搜集汇总

背景与挑战

背景概述

该数据集采集自浙江省某典型城市复杂路口，包含1小时1080p高清脱敏视频（MPEG格式）和7组连续激光雷达点云序列（BIN格式），数据由工业级设备采集，保留了空间坐标和反射强度信息。它专用于边缘计算场景的转码测试，通过硬件加速环境下的基准测试验证数据吞吐性能，旨在填补路侧边缘计算领域转码基准数据的空白，为低延迟压缩协议研究和交通数字孪生系统构建提供关键支撑。