A vision-based auto-guided vehicle system (AGVS) is introduced in the paper. Compares to the traditional cable guided vehicle

vision-based vehicle can adapt better to modern FMS (Flexible Manufacturing System). A white line marker-detecting algorithm

in which the maximum square error method is used to calculate the white line marker gray level threshold

is introduced in the paper. The vision system of the AGVS used in the paper gives deformation in the white line image

so it needs to rectify the white line image deformation before we get correct information about the white line from the image. An algorithm is proposed for this purpose. On the crossroad

the line is numbered to differentiate the different roads. Image templates matching method is used to recognize the different numbers

which represent the different roads at the crossroad. Then AGVS can use the road numbers to decide which road the car should go along.