Most methods address occluded pedestrian Re-Identification (Re-ID) by employing external auxiliary models in the feature output stage of the backbone network to locate visible appearance areas. Nevertheless, these approaches suffer from issues such as occlusion information diffusion and imprecise masks generated by external models, indicating the need for further exploration in the decoupling of pedestrian features from occlusion information. In light of these challenges, we propose an innovative algorithm called Pose-Skeleton guided Cross-attention Representation fusion (PSCR) method. Firstly, we introduce the Visible Appearance Region Attention (VARA) model designed to leverage pose information for guiding the backbone network in effectively distinguishing between occlusion information and pedestrian features at the intermediate layer. By employing a suppression strategy, the model is able to effectively suppress occlusion interference and alleviate the diffusion of occlusion information. Next, to achieve precise localization of pedestrian-specific semantic regions, a groundbreaking Skeletal Area Modeling (SAM) is proposed. Leveraging the principles of mathematical modeling and capitalizing on the efficacy of human keypoint confidence, this module generates finely-grained masks for local skeleton regions and extracts an exhaustive set of local features. Lastly, under the constraints imposed by spatial attention masks, a cross-attention mechanism is employed to fuse the features acquired from the previous two steps with local features. This fusion process results in the generation of enhanced local features that seamlessly integrate aligning high-level semantic information. Extensive experimentation demonstrates that the proposed algorithm exhibits notable performance advancements when compared to existing methodologies.

Retrieving an occluded pedestrian remains a challenging problem in person re-identification (re-id). Most existing methods utilize external detectors to disentangle the visible body parts. However, these methods are unstable due to domain bias and consume numerous computing resources. In this paper, we propose a novel and lightweight Part-based Representation Enhancement (PRE) network for occluded re-id that takes full advantages of the local correlations to aggregate distinctive information for local features without relying on auxiliary detectors. First, according to the information qualities of different body parts, we design a reasonable partition strategy to obtain the local features. Next, a Partial Relationship Aggregation (PRA) module is developed to self-mine the visibility of the body and construct a correlation matrix for collecting the information related to pre-defined classes. Following this, we propose an Inter-part Omnibearing Fusion (IOF) module that leverages the occlusion-suppressed class features to enhance the distinctiveness of the local features via feature completion and reverse fusion strategies. During the testing phase, the global and reconstructed local features are concatenated together for re-id without a complex visible region matching algorithm. Extensive experiments on occluded, partial, and holistic re-id benchmarks demonstrate the superiority of PRE over state-of-the-art methods in terms of accuracy and model complexity.