Image segmentation and scale-invariable feature extraction for level bars calibration

This paper presents a two stage method based on image segmentation and scale-invariable feature extraction for real-time level bar calibration. Firstly, images are preprocessed by averaging filtering to reduce the noise, biquadratic interpolation to increase the measuring accuracy and optimized image segmentation to improve recognition robustness. Then, a scale-invariable feature is extracted for the recognition and the measurement is carried out in sub-element level. The calculation load for the scale-invariable feature is light so the real-time calibration requirement can be guaranteed.