Image Recognition Method of Liquid Column Manometer Readings

To meet the needs of remote fluid mechanics experiment, the image processing method was proposed to read the liquid column manometer on real-time. First, image background difference operation was carried out, which used the image got at the beginning of each experiment as the background image, and the steady water head image as the target image. Second, it was proved that self-adapting iterative threshold method is effective to binary the water head difference image. Third, morphological filtering was performed on binary image for obtaining clear water head image. For eliminating the environment interference, based on the clear image, an improved string matching method was studied, which is helpful to recognize water head. According to the proportional relations between invariable inside diameter of manometer and water head image, calculated the water pressure. The experimental results show that absolute error of water pressures are in -1.0~1.0mm. The image recognition method is real-time, repeatable and feasible.