基于改进OpenPose算法的矿工危险行为识别研究

Research on Miners’Dangerous Behavior Recognition Based on Improved OpenPose Algorithm

针对现有危险行为检测中原有的人体姿态行为识别算法OpenPose存在参数量大、算力要求高的缺点,在保证准确度的情况下,提出采用MobileNet v3网络的1—18层代替OpenPose原有的VGG19网络前10层,同时采用ST-GCN时空图卷积网络完成动作的分类,实现对危险行为的识别。实验结果表明,本次改进的算法对于摔倒、危险攀爬等行为的识别准确率达到94%以上,较原有的模型准确率提升了5%以上。同时,经过对模型的轻量化改变,使得模型体积缩小、参数较少,提高了模型的运算速度。

In view of the shortcomings of OpenPose,the original human posture behavior recognition algorithm in the existing dangerous behavior detection,which has a large number of parameters andh igh computational power requirements,under the premise of enusring accuracy,it is proposed to replace the first 10 layers of OpenPose’s original VGG19 network with 1-18 layers of MobileNet v3 network,and at the same time use ST-GCN space-time map convolution network to complete the classification of actions to reazlie the recognition of dangerous behaviors.The experimental results show that the recognition accuracy of the improved algorithm fofra lling,dangerous climbing and other behaviors is more than 94%,which is more than 5%higher than the original model.At the same time,through the lightweight change of the model,the volume of the model is reduced,the parameters are less,and the calculation speed of the model is improved.