Abstract: Model segmentation has important application value in practical engineering. For the point
cloud segmentation of 3D book model, a single page extraction method based on texture is proposed.
For the 3D model books and disordered texture image acquired, the disordered texture of the model is
transformed into orderly texture firstly. Then the pixel values are converted from RGB color space to
HSI space. According to H (hue) and I (intensity), the point cloud of background model is removed.
Next, the spine line is analyzed from the distribution characteristics of the spine pixels in an orderly
texture image. The three-dimensional space equation of spine is calculated by the least squares method,
so that the space equation can be solved for the single side page extracting. Experimental results show
that the method can extract the single page stably for the 3D book model, and has higher accuracy.

Key words: page extraction, point cloud segmentation, HSI color space, book model, least square method