燕山大学学报2026，Vol.50Issue(3)：258-266,9.DOI:10.3969/j.issn.1007-791X.2026.03.008
遮挡意识增强的行人检测方法
Pedestrian detection method with enhanced occlusion awareness
摘要
Abstract
Occlusion of varying degrees and the challenges associated with modeling complex occlusion relationships are critical factors affecting pedestrian detection performance.To address these issues,an enhanced occlusion aware pedestrian detection method is proposed based on the end-to-end detection framework.First,an occlusion prior positional encoder is introduced to model occlusion relationships and provide occlusion prior information to the multi-layer decoder.Second,a multi-scale feature fusion heatmap predictor is designed to accurately capture the target center position and enhance target perception ability,thereby improving detection robustness in occluded scenes.Finally,comparative and ablation experiments are conducted on the CrowdHuman and CityPersons datasets.The experimental results demonstrate that the proposed method effectively alleviates the issues of missed detections and false positives under occlusion conditions.关键词
模式识别/遮挡检测/目标检测/遮挡意识增强/TransformerKey words
pattern recognition/occlusion detection/object detection/occlusion aware enhancement/Transformer分类
信息技术与安全科学引用本文复制引用
杨晟,张家伟,韩少杰,张世辉..遮挡意识增强的行人检测方法[J].燕山大学学报,2026,50(3):258-266,9.基金项目
国家自然科学基金资助项目(62476235) （62476235）
中央引导地方科技研发项目(216Z0301G) （216Z0301G）
河北省自然科学基金资助项目(F2023203012) （F2023203012）
