This paper proposes a new interest point extraction method based on voting strategy to ameliorate the issues of false corners and inaccurate orientation in traditional algorithms. The new method obtains the strongest feature point set without false corners through voting by the Electoral College. Interest points with high feature property meet multi-criteria in the set. In order to guarantee the accuracy orientation, this work selects key points using the same coordinate. The validity of the herewith described method is illustrated by being tested on lots of images in the cases of similarity transformation, brightness change and adding noise. Experimental results show that the method has better adaptability and robustness to noise, thanks to the more validity key points with inaccurate orientation and containing no false ones.