Previously we reported a method for displaying the shape of an object by using an array of whisker sensors. The array of whiskers was pressed against the object to be measured and the deflection of each whisker was measured. The measurements were used as inputs to a neural network to recognize the shapes. However, the previous method needed a larger number of whiskers to recognize the wider surface, because the array of whiskers has to be in full contact with the entire surface. Therefore, this study presents a new method in which a much smaller number of whiskers are dragged across the surface of the measured object to distinguish the shapes. The method is performed in the following manner. Firstly, a smaller number of whiskers are dragged across an object and each whisker deflection direction is distinguished at discrete time intervals. Next, the neural network continuously monitors the output from the whiskers and produces a result that characterizes the surface at the current scan position. Lastly, by building up a map of the classification of the entire surface, the geometry of the measured object can be displayed. This study exhibits the hardware and software required for displaying shapes, and shows examples of the results obtained from the measurements of some objects.