翻墙检测

针对校园、工地围墙等各类需要重点治安管控的室内外场所，监测是否有人员翻越围墙场景，若有及时产生告警，减小治安风险。
精度对应场景或者限制条件
1、场景限定：含有围墙的室外场景，墙体高于正常人体高度，暂不支持栅栏、栏杆，必须划定ROI（告警区域），墙体与摄像头夹角不可接近90°（70°-110°）
2、摄像头高度及角度设置：
a)较高墙体>5m，摄像头安装高度2m-5m，垂直俯仰角0-30°
b)普通墙体>2m，墙体摄像头安装高度3m-5m，垂直俯仰角20-30°
3、建议视频参数：分辨率≥1080P；码率不小于2M；帧率不小于15帧
识别对象尺寸：以1920w*1080h为基准,人体大小：>6

This dataset is designed for indoor and outdoor key security management scenarios including campus and construction site walls, to monitor personnel climbing over wall scenarios, trigger timely alarms upon detection, and mitigate security risks.

Scenario and Restriction Specifications Corresponding to Detection Accuracy:
1. Scene Limitations:
Outdoor scenarios with walls higher than the average human height; fences and railings are not supported for now. ROI (alert area) must be delineated. The angle between the wall and the camera should not be close to 90° (70°-110°).
2. Camera Height and Angle Settings:
a) For walls taller than 5m: the camera installation height ranges from 2m to 5m, with a vertical pitch angle of 0°-30°.
b) For ordinary walls taller than 2m: the camera installation height ranges from 3m to 5m, with a vertical pitch angle of 20°-30°.
3. Recommended Video Parameters:
Resolution ≥ 1080P; bitrate ≥ 2 Mbps; frame rate ≥ 15 fps.
Recognition Object Size: Taking 1920w × 1080h as the benchmark, the size of the human body should be >6.

搜集汇总

背景与挑战

背景概述

该数据集是一个针对围墙翻越行为的检测模型，适用于校园、工地等需要治安管控的场所，能够及时告警以减少风险。使用时有具体的场景、摄像头设置和视频参数要求。