From: Vehicle color classification using manifold learning methods from urban surveillance videos
Methods | Features | Classifiers | Average accuracy rates | Computational time (ms) |
---|---|---|---|---|
Baek [3] | H (36)*S (10) | SVM | 73.88 (±1.0) | 18 |
Kim [4] | H (8)*S (4)*I (4) | 1-NN | 71.04 (±1.12) | 824 |
Yang [5] | Layer 1: H (16) + S (8) | Two-layer rule-based classifier | 64.03 (±1.3) | 34 |
Layer 2: normalized RGB | ||||
Hsieh [11] | Lab + transformed RGB | GMM + two-stage SVM | 84.77 (±0.83) | 58 |
Dule [21] | HS (64) + SV (64) + ab (64) | Neural network | 76.12 (±1.41) | 1,210 |
+La (64) + Lb (64) + Gray(8) | ||||
Wu [22] | HS (256) + HV (256) + SV (256) | Two-stage SVM | 80.66 (±1.5) | 33 |
The proposed method | Six color spaces (4,608) | NFL (20) + SVM (RBF-kernel function) | 88.18 (±0.89) | 18 |
PCA reduction (200) |