Name
Abstract | ||
|---|---|---|
Bilinear models have been proposed to separate the factors from the observations for joint factors identification or translation tasks. However, the performance of existing bilinear models may degrade under challenging conditions when local image information cannot be obtained caused by occlusions or image noises. In this paper, a novel sub-pattern bilinear model (SpBM) is proposed. Different from existing bilinear models, SpBM constructs the sub-pattern bilinear model through a novel learning algorithm utilizing local patterns generated by dividing global patterns in a deterministic way. As a result, the specific factors of testing observation are identified by synthesizing the discriminative information provided by the local sub-patterns. To further improve the identification performance of SpBM, a new ridge regressive parameter estimation algorithm (RRPE) is also proposed. RRPE introduces the ridge regression into parameter estimation to stabilize the matrix inverse computation and alleviate the non-convergent cases. The proposed sub-pattern bilinear model is introduced into pose estimation of work-pieces to separate and estimate some key pose factors individually. Experimental results demonstrate the effectiveness of the proposed method. |
Year | DOI | Venue |
|---|---|---|
2012 | 10.1016/j.neucom.2011.12.012 | Neurocomputing |
Keywords | Field | DocType |
local image information,proposed sub-pattern bilinear model,local sub-patterns,local pattern,novel sub-pattern bilinear model,bilinear model,sub-pattern bilinear model,joint factors identification,estimation algorithm,pose estimation | Matrix inverse,Division (mathematics),Pattern recognition,Regression,Pose,Artificial intelligence,Estimation theory,Discriminative model,Mathematics,Computation,Bilinear interpolation | Journal |
Volume | ISSN | Citations |
83, | 0925-2312 | 0 |
PageRank | References | Authors |
0.34 | 24 | 4 |
Name | Order | Citations | PageRank |
|---|---|---|---|
| Zhicai Ou | 1 | 12 | 1.93 |
| Peng Wang | 2 | 31 | 8.02 |
| Jianhua Su | 3 | 30 | 6.44 |
| Hong Qiao | 4 | 1147 | 110.95 |