Research on Field Worker Identification Technology Under Occlusion Conditions Based on Mutual Learning
Targeting the recognition of on-site workers under occlusion conditions,this paper proposes a global feature representation method based on mutual learning.The method designs a multi-branch network architecture,where each branch independently focuses on processing images of workers in both occluded and non-occluded scenarios.Through a mutual learning mechanism,the complementary and joint improvement of feature extraction capabilities in these two extreme scenarios are achieved.Specifically,by adopting category mutual learning and feature mutual comparison strategies,not only the category consistency between occluded and non-occluded data is promoted,but also the mutual reference at the feature level is strengthened,thereby significantly enhancing the robustness and discriminability of the global feature representation.Furthermore,to better align with actual operational scenes,a data augmentation technique based on real-scenario simulation has been introduced.By generating highly realistic occlusion data,the generalization ability of the model in complex occlusion environments has been effectively improved.Experimental results show that the proposed method exhibits excellent performance in the task of recognizing on-site workers under occlusion conditions.
deep learningfield worker identificationocclusionfeature extractionrobustness
万方数据
维普
