Abstract: A new space-location vision system of cameras in multiple linear arrays is developed making full use of their high speed resolution. The space location is determined on the basic principle that some projective planes intersect with each other in non-parallel space and their locations can be solved by geometric projection. Then, a nonlinear correction is proposed by means of neural network for those cameras to enable the PSD after correction to output high-linearity location signal in a broader range. Experimental results showed that the pose measurement system for those cameras can reduce effectively the computational complexity in the space location for stereovision, and that the cameras work well at location accuracy range and sampling speed.