Water level measurement method based on deep convolutional network and random field

Abstract

The invention provides a water level measurement method based on a deep convolutional network and a random field, and the method comprises the steps: carrying out the training through a constructed water surface data set based on a proposed deep network structure; selecting a proper observation point, recording the coordinates of the characteristic pixel points and the corresponding elevations ofthe characteristic pixel points for the stable markers in the observation range, and constructing a pixel elevation interpolation function through an interpolation method; denoising and stabilizing the observed image, inputting the denoised and stabilized observed image into the trained deep convolutional network for prediction, performing preliminary segmentation according to a prediction result,and establishing a conditional probability field model; the KL divergence of the average field approximate minimization approximate distribution and the target distribution is adopted to obtain the optimal water surface segmentation, and then the elevation, namely the water level, of the pixels of the water surface region in the optimal segmentation is obtained according to the pixels of the water surface region and the elevation interpolation function. The method has the advantages that a non-contact method is adopted, and the water level is automatically monitored in real time. Compared with other contact-type measuring equipment, the method is low in layout cost and has an abnormal correction capability.

Description

technical field [0001] The invention belongs to the technical field of water level monitoring, in particular to a water level measurement method based on a deep convolutional network and a random field. Background technique [0002] Hydrological testing is an important basic work of the country. As an important part of hydrological survey, water level measurement plays an important role in the planning and management of water resources, flood control and drought relief. With the improvement of informatization, the requirements for water level measurement are gradually moving towards automation and intelligence. This creates a requirement for water level data to be collected for extended periods of time without staff being present. [0003] For water level monitoring, the commonly used means are to set up a water gauge for manual observation, and to set up a pressure sensor. At present, the widely used method for automatic monitoring is a pressure sensor, which has the prob...

AI Technical Summary

Problems solved by technology

[0004] The main purpose of the present invention is to provide a water level measurement method based on deep convolution network and random field to solve the problem of using image data to invert water level

Images