基于深度学习的污染场地作业人员着装规范性检测

为解决污染场地修复作业中缺乏统一的人员安全保障管理措施以及人工监管困难的问题，结合HSE理念提出1种基于优化Faster R-CNN的作业人员着装规范性检测算法。该方法在回归损失函数中引入L2正则项，既保证模型的泛化能力，又提高深层网络模型收敛速度。基于自建着装规范数据集（Dress Code Dataset）进行实验，评价算法检测时间和mAP等指标。结果表明:所提出的着装规范性检测算法检测时间为44 ms，mAP为88.17%，解决了传统检测算法中实时性和准确率低的问题，且模型具有更好的泛化性和鲁棒性。

Detection on normalization of operating personnel dressing at contaminated sites based on deep learning

In order to solve the problem about the lacking of unified management measures for personnel safety guarantee and the difficulty of manual supervision in the remediation operation of contaminated sites,an algorithm for the detection on the normalization of operators dressing based on the optimized Faster R-CNN was put forward combining with the HSE concept.The L2 regularization term was introduced into the regression loss function,which not only guaranteed the generalization ability of the model,but also improved the convergence speed of the deep network model.The experiments were carried out on the self built Dress Code Dataset,and the detection time,mAP and other indexes of the algorithm were evaluated.The results showed that the detection time of the proposed algorithm for the detection on dressing normalization was 44 ms,and mAP was 88.17%,which solved the problem of low real time and accuracy in the traditional detection algorithms,and the model has better generalization and robustness.