The measurement procedure was as follows. First, the
teaching pattern data for the neural networks were obtained
in 3mm intervals in the z direction. The distance between the
sensor array and the object was 200mm to 290mm. For material
identification, the pose of the object was changed in 2 degree
intervals from -10◦ to 10◦. Next, learning was performed
by the neural networks per distance. In this experiment, the
neural networks required the teaching patterns obtained in the
range of 6mm in the z direction in consideration of the error
of the distance measurement. Table II lists the number of units
in the neural networks. Learning was based on the error backpropagation
method, with the number of learning steps exceeding
10,000, or with learning continuing until the learning error
become less than 0.001. Finally, an object was measured as test
pattern. We used 31 measurement patterns for shape identification,
and 341 measurement patterns for shape identification