Answers are grounded in this paper and cite their sources.
A novel algorithm improves person re-identification accuracy in agricultural workers, highlighting automation's role in farm management.
Highlights Creatively combine person re-identification technology with agricultural scenarios. Construct a specific person re-identification dataset of agricultural workers. A novel person re-identification algorithm proposed exhibits good performance. Abstract. To achieve sustainable agricultural development, effective supervision of agricultural workers on farms is crucial. Traditional manual recording methods suffer from inefficiencies in data management. Therefore, automating record-keeping from video footage represents a viable approach to enhancing management efficiency. Addressing the challenge of identifying individuals in video, person re-identification has emerged as a pivotal research direction in computer vision. This article first constructs a unique person re-identification dataset, FWreID, tailored specifically to agricultural workers, comprising 4,464 images of 240 workers. Subsequently, a multi-feature fusion method for person re-ID is proposed. This method incorporates a global branch, a local branch, and an appearance-contour fusion network. During convolutional neural network feature extraction, contour features are introduced, and an attention-based fusion network branch is devised to enhance the network's ability to capture contour information under varying poses. To address occlusion and extract distinctive features from individuals, a local feature fusion module is designed, contributing to the overall multi-feature fusion person re-ID model. By integrating global and local information and innovatively designing the appearance-contour fusion network, this method overcomes limitations such as single-view or missing information typical in traditional approaches, significantly improving the accuracy and robustness of agricultural worker identification. Experimental results demonstrate that the proposed multi-feature fusion method achieves superior performance compared to other person re-ID approaches, with rank-1 accuracy of 95.3% and mAP of 87.6% on the FWreID dataset. Furthermore, it shows strong robustness with rank-1 accuracies of 95.6% and 89.1%, and mAP of 88.4% and 79.4% on the public datasets Market1501 and DukeMTMC-reID, respectively. This advancement is pivotal for digitizing and standardizing the management of farm workers by agricultural enterprises, thereby promoting the further automation and smart transformation of traditional agriculture.
No takes yet. Share an insight, caveat, or question.
Xiao et al. (2025) studied this question.
Synapse has enriched 10 related papers. Consider them for comparative context: