The prototype system has a good linearity of
1.17%. Fig. 10 shows the standard deviations of the distance
detection versus the distance in the z direction, when the objects
pose were from -10◦ to 10◦. In this way, the accuracy ofthe distance detection is easily influenced by the pose of measured objects. The proposed method need to select a suitable
neural network based on the detected distance. Therefore, the
standard deviation of the distance detection was important. In
this experiment, The standard deviation was lower than 2.5mm
up to 280mm in the z direction.