Water gauge intelligent identification method based on deep learning

Abstract

The invention relates to a water gauge intelligent identification method based on deep learning. The method comprises the following steps: 1, shooting a water gauge in a river channel by using a video shooting device so as to obtain a real-time image of the water gauge in a natural scene, and transmitting image data to a client server by a central control module; 2, enabling the central control module to input the water gauge image to an integrated water gauge identification network on a client server for training; 3, enabling the central control module to input the water gauge image to a trained integrated water gauge recognition network for recognition, and enabling the integrated water gauge recognition network to calculate the accurate water level height through the upper end of the water gauge and the edge of the water line. Through the method, the problems that the water gauge image shot by the video shooting device is influenced by illumination, the image noise is much, the target is not obvious and the shot image is not clear can be effectively reduced, and the problems that a large number of existing deep learning networks need a large number of training samples, are complex and are difficult to be directly applied to the field of water conservancy are solved.

Description

technical field [0001] The invention belongs to the field of machine vision detection, and in particular relates to an intelligent identification method for water gauges based on deep learning. Background technique [0002] The establishment of video surveillance systems in rivers, lakes and other places can greatly accelerate the pace of water conservancy information construction. Generally speaking, the traditional water level measurement methods mainly include the installation of water gauges for visual readings and the use of sensors to automatically collect analog quantities related to water levels and then convert them into water level quantities. Among them, the visual measurement method has low efficiency, poor timeliness, and cannot be read under harsh conditions, while the sensor measurement method is costly, difficult to maintain, and greatly affected by the environment. Through the visual processing technology, the situation of the reservoir can be observed in r...

AI Technical Summary

Problems solved by technology

[0004] For this reason, the present invention provides a water gauge intelligent recognition method based on deep learning, which is used to overcome the traditional machine vision water gauge recognition algorithm in the prior art, which is obviously affected by light, has a lot of image noise, and the target is not obvious. A large number of existing deep learning networks that are not clear and require a large number of training samples and the network is complex, so it is difficult to directly apply to the field of water conservancy

Images