微型电脑应用2025，Vol.41Issue(2)：74-77,4.
基于多目标识别分析的智能作业监控算法设计与仿真
Design and Simulation of Intelligent Job Monitoring Algorithm Based on Multi-objective Recognition Analysis
摘要
Abstract
In order to improve the safety control level of substation engineering construction sites,an intelligent video monito-ring algorithm is designed based on usage scenarios such as video capture,environmental monitoring,large screen display,vid-eo storage servers,and intelligent supervision platform servers.The proposed algorithm can be abstracted as a deep neural net-work that integrates convolutional computation,pooling computation,and Softmax classifier.It can quickly and accurately re-place low-level feature maps with high-level ones,thereby realize the recognition of typical violations at the job site.In addi-tion,considering the timeliness of capturing video frames,the algorithm introduces a bidirectional long short-term memory net-work to replace the full connection layer in the traditional deep network,which improves the efficiency of feature extraction while avoiding the gradient disappearance of network training.The field deployment test results show that the designed algo-rithm can not only achieve intelligent detection of multiple objects,but also achieve an overall mAP of 96.23％,which is about 3.74 percentage points higher than traditional CNN.关键词
深度学习/目标检测/算法设计/Bi-LSTM/变电站/违章识别Key words
deep learning/object detection/algorithm design/Bi-LSTM/substation/violation identification分类
信息技术与安全科学引用本文复制引用
李俊,金燕云,王坚,冯艳..基于多目标识别分析的智能作业监控算法设计与仿真[J].微型电脑应用,2025,41(2):74-77,4.基金项目
国网湖北省电力有限公司基建数字化重点项目(HB2019A576) （HB2019A576）
