Research On Remote Water Level Monitoring System Based On Image Recognition

The safety and the quality of water resources has currently become a significant factor to influence the development of society and economy. And in our country, catastrophic floods occur more frequently. However, at present, many domestic hydrological stations still apply manual method to monitor the water level and this method severely pose threats towards the safety of monitors, moreover, the data measured with this method is quite inaccurate and not real-time.For these problems, this paper combines image recognition technology and wireless communication technology together to develop a remote water level monitoring system. This system applies image recognition technique to recognize water level. Then the water level data is accurately transmitted to the monitoring end through the wireless communication network, therefore, the remote monitor of the water level is realized. The main research contents of this paper are as follows:1. A detailed introduction of the overall design of the remote water level monitoring system is provided, which includes smart phone image collection, the programming of the communication of wireless communication module and computer serial interface.2. Two kinds of water gauge localization algorithms are presented, one is based on image connected domain analysis and mathematical morphology, which extracts image connected domain characteristics and processes prior restrictions. The other is based on the color of water gauge scale, which adopts color segmentation method for water gauge location.3. This paper investigates a common method for tilt angle detection, and put forward a tilt correction methods based on the minimum connected domain rectangular area. Through experiments, this algorithm turns out to perform with a high accuracy and meanwhile guarantees the system’s requirement for real-time property; therefore, it is applicable to tilt correction of gauge.4. A method for water level recognition is put forward, which recognize water level the scale lines by means of the gray transform, horizontal edge detection and k-means clustering algorithm.