校园安全与异常行为识别数据集

本数据集基于校园多场景监控视频，融合刷卡记录、音频事件等信息，经过深度加工形成结构化行为知识库。原始视频数据被转化为包含行为序列、场景上下文、风险等级的智能考评样本，每条记录涵盖行为分层标签（如打架斗殴/心理异常/安全隐患）、时空轨迹信息（前一状态/后一状态）、多视角关联及对抗样本标记。数据集采用事件演化逻辑编排，将碎片化的监控画面转化为可推理的行为图谱，实现了从原始视频到智能决策信息的本质转变，为校园安全管理的智能化考评提供了可直接操作的数据支撑。This dataset is constructed based on multi-scenario campus surveillance videos, fused with information such as swiping records and audio events, and subjected to in-depth processing to form a structured behavioral knowledge base. The original video data is converted into intelligent evaluation samples containing behavior sequences, scene context and risk levels. Each record covers hierarchical behavior tags (such as fighting, psychological abnormalities and safety hazards), spatiotemporal trajectory information (previous state and subsequent state), multi-view association and adversarial sample markers. The dataset is organized according to the logic of event evolution, transforming fragmented surveillance footage into inferable behavioral graphs, achieving a fundamental shift from raw video to intelligent decision-making information, and providing directly operable data support for intelligent evaluation of campus safety management.

搜集汇总

背景与挑战

背景概述

该数据集是一个针对校园安全管理的多源信息融合数据集，基于监控视频、刷卡记录和音频事件等数据，经过深度加工形成结构化行为知识库。它将原始视频转化为包含行为序列、场景上下文和风险等级的智能考评样本，每条记录涵盖行为分层标签、时空轨迹信息及多视角关联，支持事件演化逻辑和可推理的行为图谱构建。该数据集旨在实现从视频数据到智能决策信息的本质转变，为校园安全管理的智能化考评提供可直接操作的数据支撑。