车载视频监控异常行为识别数据集

本数据集面向车载视频监控与道路运输安全运营场景，以货车和客车驾驶过程中的异常行为片段为最小数据粒度，对车载监控终端输出的视觉识别特征、驾驶操作状态、车辆运行状态、算法识别结果及人工复核结论进行结构化归集。数据覆盖疲劳驾驶、打电话、吸烟、分心侧视、打哈欠、双手离盘、未系安全带七类典型异常行为，并区分重型半挂、城配轻卡、冷链厢式、危化运输、旅游客车、城际客运、城市公交和通勤班车等车辆细分类型。数据字段既包含采集日期、匿名终端、匿名行程、监控时段、摄像头视角、路段类型、行驶速度等业务维度，也包含面部遮挡比例、眼睛闭合比例、头部偏转角、嘴部张开比例、方向盘离手时长、手持物体置信度等视觉及动作特征。在基础识别结果之上，进一步生成异常行为强度分、视觉置信修正值、连续风险累计指数、行驶风险综合分，此数据集旨在为驾驶安全管理、驾驶员行为分析、智能监控算法研发及道路交通安全研究提供坚实可靠的数据支持。This dataset is tailored for in-vehicle video monitoring and road transportation safety operation scenarios, taking abnormal behavior segments during the driving of freight and passenger vehicles as the minimum data granularity. It structurally collects visual recognition features, driving operation status, vehicle operating status, algorithm recognition results and manual review outputs from in-vehicle monitoring terminals. The dataset covers seven typical abnormal behaviors: fatigue driving, phone calling, smoking, distracted side-glancing, yawning, hands-off-steering wheel, and failure to wear a seatbelt, and differentiates vehicle subtypes including heavy-duty semi-trailers, urban distribution light trucks, cold-chain box trucks, hazardous chemical transport vehicles, tourist coaches, intercity passenger coaches, urban buses and commuter buses. The data fields include both business dimensions such as collection date, anonymous terminal, anonymous trip, monitoring period, camera angle, road segment type and driving speed, as well as visual and motion features including facial occlusion ratio, eye closure ratio, head deflection angle, mouth opening ratio, steering wheel off-hand duration and handheld object confidence. Based on the basic recognition results, it further generates abnormal behavior intensity score, visual confidence correction value, continuous risk cumulative index and comprehensive driving risk score. This dataset aims to provide solid and reliable data support for driving safety management, driver behavior analysis, intelligent monitoring algorithm development and road traffic safety research.

搜集汇总

背景与挑战

背景概述

该数据集针对货车和客车驾驶中的疲劳驾驶、打电话、吸烟等七类异常行为，整合了车载监控终端的视觉特征、驾驶操作状态及车辆运行数据，并按重型半挂、城配轻卡、旅游客车等车辆类型进行细分。数据字段涵盖采集日期、终端行程、摄像头视角、速度等业务维度，以及面部遮挡、眼睛闭合、头部偏转等视觉特征，并生成异常行为强度分和行驶风险综合指数，旨在支撑驾驶安全管理、驾驶员行为分析及智能监控算法研发。