Visual Recognition of Objects with Sparse Texture and Specular Reflectivity

摘要

Appearance-based object recognition with traditionalrn2-D cameras requires the extraction of image features thatrncan be used for classification. Scale Invariant FeaturernTransforms, (SIFT) [1], Speeded Up Robust Features,rn(SURF) [2], and Histograms of Oriented Gradientsrn(HOG) [3] are examples of features that are used forrnclassification in appearance-based object recognitionrnmethods. These features are usually determined from thernlocal image gradient with the result that they requirernsufficient image texture to classify objects robustly.