Bridge side falling behavior identification method based on depth understanding

• Main Authors: CHENG XIAOGANG, NI JIE, XU FENGLEI, ZHANG BO, WANG ZHAOBIN, ZHU JIAXIANG, CAI CONGCONG, GAO BO
• Format: Patent
• Language: Chinese; English
• Published: 01.04.2022
• Subjects: CALCULATING; COMPUTING; COUNTING; PHYSICS

The invention discloses a bridge side falling behavior identification method based on deep understanding, and the method employs a camera which monitors a large bridge all the time to capture a signal that a person falls from a bridge side, and gives out an alarm signal, thereby enabling the person to be rescued in time. According to the system, a computer vision algorithm is embedded in a camera on a river bridge, and the system comprises a personnel railing crossing behavior monitoring module, a personnel falling monitoring module, a falling water spray monitoring module and a personnel floating detection and rescue area prediction module. According to the system, whether a person climbs over a river-crossing bridge handrail and falls from the bridge side is judged through cross validation of the first three modules, and if the person is detected, an alarm can be given and rescue can be called in time, so that the best rescue time is prevented from being missed; the rear two modules are used for predicting