MRRF法を用いた複雑背景下における人物の姿勢認識(バイオメディカル・画像処理その4)

抄録

In this paper, we propose a technique for extracting a human region from images and for recognizing human postures. We employ a background subtraction technique using texture information and gray scale for extracting a human region and a technique for recognizing a 3-D object from a 2-D image. Many extraction techniques must be performed in the outdoors and need thresholds tuned by manual operations and many recognition techniques need extraction of specific characteristics or modeling. Since the proposed technique does not necessitate manual operations, it can perform the recognition from extraction automatically. The proposed technique obtained satisfactory results by the experiment performed in the outdoor environments.