Firstly

a fisheye lens with a wide-angle field of view is used to capture images containing multiple pointer meters. Then three steps of pre-processing operation are carried out for multi-meters image. Firstiy

image graying is carried out. secondly

the barrel distortion of fish-eye lens is eliminated; thirdly

the image distortion caused by the non-parallel of lens plane and dial plane is eliminated

and the multi-meters image are divided into several single-meter images. Finally

reading recognition of each single-meter image in turn

the processes include dial extraction

pointer detection

and reading calculation. The experimental results show that in the case of the segmented single-meter image quality is not high

the satisfactory reading results can still be obtained.