摘要
以往的造纸车间监控视频异常识别方法由于仅对数据进行了脱敏处理,导致数据识别准确率不高。因此,设计了基于多特征融合的造纸车间监控视频异常识别方法。首先对数据进行预处理,通过脱敏数据、数据填补和颠簸去除,保证数据的完整性,根据动态数据的特征,提取数据轨迹特征,基于多特征融合计算异常数据调度函数,构建监控视频异常识别模型,在上述基础上,通过判断当前数据的状态,实现数据的异常识别。通过上述的设计,完成对造纸车间监控视频异常识别方法的设计。在仿真实验中,和以往的造纸车间监控视频异常识别方法相比,设计的基于多特征融合的造纸车间监控视频异常识别方法识别准确率平均值为98.7%,能更加准确地识别异常数据。
The previous methods for identifying abnormal surveillance videos in paper mills only performed data desensitization processing,resulting in low accuracy in data recognition.Therefore,a paper workshop surveillance video anomaly recognition method based on multi feature fusion was designed.Firstly,the data is preprocessed through desensitization,data filling,and turbulence removal to ensure data integrity.Based on the characteristics of dynamic data,trajectory features are extracted from the data,and abnormality scheduling functions are calculated based on multi-feature fusion to construct a monitoring video anomaly recognition model.Based on the above,the current state of the data is judged to achieve anomaly recognition.Through this design,an anomaly recognition method for monitoring videos in the papermaking workshop is completed.In simulation experiments,compared with previous methods for identifying food anomalies in food surveillance videos from paper workshop surveillance,the designed paper workshop surveillance video anomaly recognition method based on multi-feature fusion has achieves an average recognition accuracy of 98.7%,enabling more accurate identification of abnormal data.
作者
胡茂伟
HU Maowei(Tsinghua University Shenzhen International Graduate School,Shenzhen 518055,China)
出处
《造纸科学与技术》
2023年第4期13-16,20,共5页
Paper Science & Technology
关键词
多特征融合
造纸车间
监控视频
监控异常识别
方法设计
multi feature fusion
paper workshop
surveillance video
monitoring anomaly identification
method design
