基于SENet-SSD的水电厂人员作业安全行为识别方法研究

On the Identification of Safety Operation Behaviors of Personnel in Hydropower Plants based on SENet-SSD

为提高水电厂人员安全行为管理水平,降低安全风险,提出了一种基于单框多目标检测算法(Single Shot MultiBox Detector, SSD)的人员安全行为图像识别模型。该模型采用Squeeze-and-Excitation Networks(SENet)对SSD模型的通道关系建模,对通道特征的重要程度进行优化,提升模型性能。通过实验得出,SENet-SSD模型的精确度(mAP)可达91.6%,有效提高了是否穿戴安全帽的识别精度,对水电厂的人员安全行为管控能力提升具有很好的支撑作用。

To improve the personnel safety behavior management level of hydropower plants and reduce the safety risks, an image recognition model of personnel safety behavior based on single shot multi-box detector(SSD) is proposed. The squeeze-and-excitation networks(SENet) is adopted to model the channel relationships in the SSD model and optimize the importance of the channel features, thus improving the model performance. Experimental results show that the accuracy(mAP) of the SENet-SSD model can reach 91.6%. It effectively improves the recognition accuracy of helmet wearing, and provides a strong support in improving the personnel safety behavior management ability of hydropower plants.