Scene recognition with majority voting among sub-section levels

2016 
In this paper, scene recognition problem, which is a frequently-studied field of computer vision, is tackled. Proposed algorithm utilizes bag of words (BoW) method along with considering sub-segments in the image during classification. For this purpose, the image is represented in three sub-segment levels where the image is divided into equal sized sub-segments at each level. The number of sub-segments are increased as the sub-segment level is increased and each sub-segment at each level is classified. During classification, responses of different sub-segment levels to classifier is considered with a major voting policy. The experiments are made on a database that contains approximately 4500 samples of scene images with dictionary sizes of 50, 100, 200, 300 and different sub-segment levels. The results show that, the proposed method achieves 71.83% accuracy and the sub-segment major voting increases the performance by % 1 according to the non-major voting case.
    • Correction
    • Source
    • Cite
    • Save
    • Machine Reading By IdeaReader
    0
    References
    0
    Citations
    NaN
    KQI
    []