A graph construction method using LBP self-representativeness for outdoor object categorization

In this paper, we introduce a new graph construction algorithm that is useful for many semi-supervised learning tasks. Unlike the main stream for graph construction, our proposed data self-representativeness approach simultaneously estimates the graph structure and its edge weights through sample coding. Compared with the recent ℓ1 graph that is based on sparse coding, our proposed objective function has a closed-form solution and thus is more efficient than the iterative schemes deployed for solving the sparse coding problem. Our proposed method is inspired by the recent coding scheme “Weighted Regularized Least Square” (WRLS) proposed for improving the Sparse Representation Classifier. aper has two main contributions. Firstly, we introduce a Two Phase Weighted Regularized Least Square (TPWRLS) graph construction that is based on self-representativeness of data samples. A key element of the proposed method is the second phase of coding that allows data closeness or locality to be naturally incorporated by solving a coding over some automatically selected relevant samples and by reinforcing the individual regularization terms according to the first phase coefficients. Secondly, the obtained data graph is used, in a semi-supervised context, in order to categorize detected objects in driving/urban scenes using Local Binary Patterns as image descriptors. The experiments show that the proposed method can outperform competing methods.