Name
Abstract | ||
|---|---|---|
The human saliency feature has been increasingly used for person re-identification across non-overlapping cameras but is deficient in retaining the minor features of the salient region, thus resulting in matching accuracy decline. To address this challenge, we first propose to extract optimal regions from pedestrian images that contain high intra-region feature similarity. Subsequently, by computing the saliency of each region, we choose the most salient region, which contains not only saliency features but also minor features, to represent the corresponding pedestrian. Finally, by formulating the competitive matching as hypothesis in a matching game, we obtain the most suitable set of matching by iteratively computing the payoff of each hypothesis. We evaluate our scheme on three widely used public datasets, and experimental results verify the advantage of our proposed algorithm, which outperforms previous representative methods with a matching ratio of 10.8%. |
Year | DOI | Venue |
|---|---|---|
2018 | 10.1007/s11042-017-5541-9 | Multimedia Tools Appl. |
Keywords | Field | DocType |
Re-identification, Salient region, Matching game, Region match | Computer vision,Pedestrian,Matching game,Pattern recognition,Computer science,Salience (neuroscience),Artificial intelligence,Stochastic game,Salient | Journal |
Volume | Issue | ISSN |
77 | 16 | 1380-7501 |
Citations | PageRank | References |
0 | 0.34 | 23 |
Authors | ||
4 | ||
Name | Order | Citations | PageRank |
|---|---|---|---|
| Tiezhu Li | 1 | 0 | 0.68 |
| Lijuan Sun | 2 | 118 | 20.41 |
| Chong Han | 3 | 314 | 32.63 |
| Jian Guo | 4 | 221 | 27.43 |