基于AlphaPose与REBA的手工搬运作业姿势风险评估方法*

为有效预防职业性肌肉骨骼疾患(WMSDs)，提出基于计算机视觉技术与快速全身评估(REBA)结合的作业姿势评估方法，以手工搬运作业为例，通过人体姿态识别AlphaPose模型检测作业姿势，利用提取的各关节点空间位置信息计算身体各部位关节角度；利用SG滤波法处理身体遮挡、光线等对数据的影响，有效保留数据变化信息的同时消除数据突变；根据身体关节角度范围与危害程度对应关系自动评判作业姿势的风险大小。研究结果表明:作业姿势评估方法可通过采集的视频数据自动评估搬运作业的姿势危险性，降低WMSDs风险，通过分析不同下蹲取物姿势风险，可论证作业姿势评估方法的可行性，为及时改善搬运作业中的不安全动作提供参考。

Risk assessment method on work posture of manual handling based on AlphaPose and REBA

In order to effectively prevent the work-related musculoskeletal diseases (WMSDs),an assessment method of work posture based on the combination of computer vision technology and rapid entire body assessment (REBA) was proposed.Taking the manual handling work as an example,firstly,the human posture recognition AlphaPose model was used to detect the work posture,and the extracted space position information of each joint point was used to calculate the joint angle of each part of the body.Secondly,the SG filtering method was used to process the influence of body occlusion,light,etc.on data,so as to effectively retain the information of data change while eliminating the data mutation.Finally,the risk of work posture was automatically evaluated according to the corresponding relationship between the range of joint angle and the degree of harm.The results showed that the assessment method of work posture could automatically evaluate the posture hazard of handling work through the collected video data,reduce the risk of WMSDs,experimentally analyze the risk of different squatting fetch postures,and verify the feasibility of the assessment method of work posture,thus provide basis for timely improve the unsafe actions in handling work.