Hand Trajectory Extraction Of Human Assembly Based On Multi-Leap Motion

Because 3C products have the characteristic of complex assembly process and frequent change,at present,the assembly,testing and packaging of 3C industry are still completed by manual.In order to improve the automation of 3C products in the assembly process,if different robot structures are designed according to different processes,on the one hand,the professional requirements of designers are too high,on the other hand,it is not conducive to the flexible application of robots in 3C assembly.In order to solve these problems,a multi-sensor data acquisition system is designed to extract the sixdimensional(position and attitude)motion trajectory of the hand movement of the assembly site workers.It lays a foundation for the automatic design and reappearance of the 3C assembly robot,which is beneficial to the rapid popularization of the robot in the 3C assembly.In order to automatically collect the actual assembly movements of 3C production site workers and meet the needs of lightweight,portable and non-contact measurement,this paper proposes a low-cost 3D sensor based on Leap Motion to track the sixdimensional(position and attitude)motion trajectory on the assembly action of field workers.The multi-sensor data fusion is used to solve the occlusion problem of a single sensor in the motion track tracking process,and the tracking accuracy of the assembly action in the target field of vision is improved at the same time.The performance of the Leap Motion sensor was evaluated by a set of static experiments.The plastic hand model is used to simulate the human hand,and 98 reference positions are selected in the visual field of the controller.The repositioning accuracy of observation points is taken as the evaluation index,and the accuracy of repeated positioning in most visual field of sensors in static measurement is within 1 mm,which meets the requirements of the subject.By circularly searching in the visual field of Leap Motion,a regular cuboid space is defined as the best workspace for a single sensor in the visual field,which meets 3C assembly requirements.Considering the performance limitation of the single Leap Motion Controller,this paper uses three Leap Motions to form a multi-sensor data acquisition system,to make up for the shortcoming of a single sensor system,that losing the recognition accuracy of a single sensor under the condition of palm occlusion or overlap.At the end of this paper,the three Leap Motions were sub-regionally calibrated by ICP algorithm,and the five most accurate points were found on the palm of the hand for calibration.The optimal calibration error was within 1 mm.Addressing the uncertainty of data in multi-sensor system,a multi-sensor data fusion algorithm based on principal component analysis(PCA)is adopted.By introducing a support degree matrix and adding palm constraints,the invalid data in the sensor observations are eliminated before fusion,and the effective observation values of the sensors are fused by principal component analysis(PCA).The experimental results show that this method has better fusion effect.