The invention provides an off-site law violation automatic discrimination method based on video deep learning. A law violation picture can be automatically located, the workload of manual discrimination is reduced, the working efficiency is improved, the evidence obtaining means is simplified, and the dispute rate of evidence obtaining is reduced. In the technical scheme of the invention, invalidscene pictures in video images are firstly recognized based on an invalid scene recognition sub-model, then the remaining pictures are put into a law violation event recognition sub-model to carry outtraffic law violation scene picture recognition, and law violation evidence obtaining data and suspected law violation video data are finally confirmed manually.
本发明提供一种基于视频深度学习的非现场违法自动甄别方法,其可以自动定位违法图片,降低了人工甄别的工作量,提高了工作效率,简化了取证手段,且降低了取证的争议率。本发明技术方案中,基于无效场景识别子模型,先将视频图像中的无效场景图片识别出来,然后将剩余图片投入到违法事件识别子模型中,进行交通违法现场图片识别,由人工对违法取证数据、疑似违法视频数据进行最终确认。
Off-site law violation automatic discrimination method based on video deep learning
一种基于视频深度学习的非现场违法自动甄别方法
2021-03-16
Patent
Elektronische Ressource
Chinesisch
Airport departure layer violation vehicle automatic identification method based on deep learning
Europäisches Patentamt | 2022
|Vehicle violation detection method based on deep learning
Europäisches Patentamt | 2022
|Deep learning based vehicle violation detection system
IEEE | 2021
|DEEP LEARNING BASED TRAFFIC VIOLATION ENFORCEMENT SYSTEM AND METHOD
Europäisches Patentamt | 2020
|Deep-learning based helmet violation detection system
IEEE | 2021
|