Aiming at the problems of low efficiency and limited accuracy of manual reading of pointer type pressure gauges， a Halcon-based machine vision technology is proposed to realize the automatic reading recognition of pointer type pressure gauges through the combination of image processing， polar coordinate transformation， projection transformation， image segmentation， extraction of pointer features， and geometric calculation. The results show that the tested accuracy of the method reaches 97.5% and the average reading time is 0.31 s. Compared with the inefficiency of the traditional human eye reading method， the method has high recognition accuracy and the reading time is improved by more than 10 times， which makes up for the shortage of manual interpretation. This method is simple， fast and conducive to improving the efficiency and the accuracy of the readings， promising a broad application prospect.