Visual functions for bin picking are realized fast by utilizing a sparse range image. The functions are composed of (1) detection of a necessary part from a lot of parts piled on each other, and (2) measurement of its three dimensional position and orientation. Framework of the vision system is discussed on the assumption that a shape model of the necessary part is given. Method of fast segmentation of a sparse range image is presented for the part's detection. Four kinds of shapes: polyhedra, cylinder, cone and sphere are chosen as important shapes for industrial bin-picking, and for each shape, examination if a segment of a sparse range image is the shape, and measurement of its three dimensional position and orientation are shown. Experiments with a real fast range sensor and a personal computer show that the presented vision system is fast: total process time is less than 1.5 s for multiple piled cylinders.