Focus on the Visible Regions: Semantic-Guided Alignment Model for Occluded Person Re-Identification
The occlusion problem is very common in pedestrian retrieval scenarios. When persons are occluded by various obstacles, the noise caused by the occluded area greatly affects the retrieval results. However, many previous pedestrian re-identification (Re-ID) methods ignore this problem. To solve it, w...
| Main Authors: | , , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
MDPI AG
2020-08-01
|
| Series: | Sensors |
| Subjects: | |
| Online Access: | https://www.mdpi.com/1424-8220/20/16/4431 |
| _version_ | 1797559450758807552 |
|---|---|
| author | Qin Yang Peizhi Wang Zihan Fang Qiyong Lu |
| author_facet | Qin Yang Peizhi Wang Zihan Fang Qiyong Lu |
| author_sort | Qin Yang |
| collection | DOAJ |
| description | The occlusion problem is very common in pedestrian retrieval scenarios. When persons are occluded by various obstacles, the noise caused by the occluded area greatly affects the retrieval results. However, many previous pedestrian re-identification (Re-ID) methods ignore this problem. To solve it, we propose a semantic-guided alignment model that uses image semantic information to separate useful information from occlusion noise. In the image preprocessing phase, we use a human semantic parsing network to generate probability maps. These maps show which regions of images are occluded, and the model automatically crops images to preserve the visible parts. In the construction phase, we fuse the probability maps with the global features of the image, and semantic information guides the model to focus on visible human regions and extract local features. During the matching process, we propose a measurement strategy that only calculates the distance of public areas (visible human areas on both images) between images, thereby suppressing the spatial misalignment caused by non-public areas. Experimental results on a series of public datasets confirm that our method outperforms previous occluded Re-ID methods, and it achieves top performance in the holistic Re-ID problem. |
| first_indexed | 2024-03-10T17:45:37Z |
| format | Article |
| id | doaj.art-17466c4b040445338cf0da06ffc3ac4f |
| institution | Directory Open Access Journal |
| issn | 1424-8220 |
| language | English |
| last_indexed | 2024-03-10T17:45:37Z |
| publishDate | 2020-08-01 |
| publisher | MDPI AG |
| record_format | Article |
| series | Sensors |
| spelling | doaj.art-17466c4b040445338cf0da06ffc3ac4f2023-11-20T09:31:04ZengMDPI AGSensors1424-82202020-08-012016443110.3390/s20164431Focus on the Visible Regions: Semantic-Guided Alignment Model for Occluded Person Re-IdentificationQin Yang0Peizhi Wang1Zihan Fang2Qiyong Lu3Key Laboratory for Information Science of Electromagnetic Waves, Research Center of Smart Networks and Systems, School of Information Science and Technology, Fudan University, Shanghai 200433, ChinaKey Laboratory for Information Science of Electromagnetic Waves, Research Center of Smart Networks and Systems, School of Information Science and Technology, Fudan University, Shanghai 200433, ChinaKey Laboratory for Information Science of Electromagnetic Waves, Research Center of Smart Networks and Systems, School of Information Science and Technology, Fudan University, Shanghai 200433, ChinaKey Laboratory for Information Science of Electromagnetic Waves, Research Center of Smart Networks and Systems, School of Information Science and Technology, Fudan University, Shanghai 200433, ChinaThe occlusion problem is very common in pedestrian retrieval scenarios. When persons are occluded by various obstacles, the noise caused by the occluded area greatly affects the retrieval results. However, many previous pedestrian re-identification (Re-ID) methods ignore this problem. To solve it, we propose a semantic-guided alignment model that uses image semantic information to separate useful information from occlusion noise. In the image preprocessing phase, we use a human semantic parsing network to generate probability maps. These maps show which regions of images are occluded, and the model automatically crops images to preserve the visible parts. In the construction phase, we fuse the probability maps with the global features of the image, and semantic information guides the model to focus on visible human regions and extract local features. During the matching process, we propose a measurement strategy that only calculates the distance of public areas (visible human areas on both images) between images, thereby suppressing the spatial misalignment caused by non-public areas. Experimental results on a series of public datasets confirm that our method outperforms previous occluded Re-ID methods, and it achieves top performance in the holistic Re-ID problem.https://www.mdpi.com/1424-8220/20/16/4431deep learningperson re-identificationocclusionsemantic segmentationfeature fusion |
| spellingShingle | Qin Yang Peizhi Wang Zihan Fang Qiyong Lu Focus on the Visible Regions: Semantic-Guided Alignment Model for Occluded Person Re-Identification Sensors deep learning person re-identification occlusion semantic segmentation feature fusion |
| title | Focus on the Visible Regions: Semantic-Guided Alignment Model for Occluded Person Re-Identification |
| title_full | Focus on the Visible Regions: Semantic-Guided Alignment Model for Occluded Person Re-Identification |
| title_fullStr | Focus on the Visible Regions: Semantic-Guided Alignment Model for Occluded Person Re-Identification |
| title_full_unstemmed | Focus on the Visible Regions: Semantic-Guided Alignment Model for Occluded Person Re-Identification |
| title_short | Focus on the Visible Regions: Semantic-Guided Alignment Model for Occluded Person Re-Identification |
| title_sort | focus on the visible regions semantic guided alignment model for occluded person re identification |
| topic | deep learning person re-identification occlusion semantic segmentation feature fusion |
| url | https://www.mdpi.com/1424-8220/20/16/4431 |
| work_keys_str_mv | AT qinyang focusonthevisibleregionssemanticguidedalignmentmodelforoccludedpersonreidentification AT peizhiwang focusonthevisibleregionssemanticguidedalignmentmodelforoccludedpersonreidentification AT zihanfang focusonthevisibleregionssemanticguidedalignmentmodelforoccludedpersonreidentification AT qiyonglu focusonthevisibleregionssemanticguidedalignmentmodelforoccludedpersonreidentification |