Robot Assembly Action Recognition and Job Standardization Evaluation Method Based on Computer-Aided Manufacturing

With the continuous progress of science and technology, industrial robots are increasingly used in manufacturing. To improve the action recognition accuracy and job standardization evaluation of human-robot collaboration in computer-aided manufacturing, this study utilizes the spatio-temporal extraction method with the improved you only look once version 4 algorithm and recognizes assembly action. Moreover, the job standardization evaluation method is used to make the robot better understand the operation intention of the personnel. The outcomes revealed that the improved YOLOv4 algorithm achieved a maximum recognition accuracy of 95.49%, a recall rate of 94.64%, and an F1 value of 92.60% in assembly action recognition. The assembly action extraction accuracy was greater than 0.90, and the similarity between background and actions before and after assembly action segmentation reaches 0.95 and 0.93, respectively. It indicated that this method could effectively preserve temporal features. Under noise interference, the standardized matching results of assembly operations remained stable, with a final matching degree of 96.48%. In the assembly operation chain, this method reduced the total assembly time to 24.97 s, which was 9.16%, 26.47%, and 33.97% shorter than other algorithms, respectively. In summary, the computer-aided manufacturing- based robot assembly action recognition and job standardization evaluation methods are effectively studied. Job standardization evaluation method can effectively improve the accuracy and manufacturing efficiency of computer-aided manufacturing assembly robot action recognition, and provide support for human-robot collaboration in the assembly field.