Meter digital identification is crucial for intelligent meter application. Current methods have lower recognition ratio at presence of tilt angle and half-digits. In order to solve this problem, the hardness feature parameters are introduced to measure anti-deformation ability on the direction of the projection target area. A number identification method based on the combination of digital structure feature and statistical feature is presented. By analysising different height and different angle of the figures acquired from Meter, the template library of digital top-down and bottom-up hardness characteristics is established. Weights are set according to the different importance of characteristics and the digital recognition is realized by using the method of weighted feature matching algorithm. Experiments prove that the algorithm is simple and efficient to words or words in half and has strong fault-tolerant with rotation and distortion.