Research On Water Level Scale Recognition Based On Digital Image Processing

The traditional water level detection technology has certain requirements for the environment due to the characteristics of their measuring equipment which has complex installation with a lot of manpower and resources.With the development of modern digital image processing and computer vision,they have been widely used in many fields such as medicine,transportation,biology,military and so on.With the camera to capture the water level scale image,processing the image to identify water level information can greatly reduce the complexity of installation and maintenance.Therefore,this article makes full use of the priori information which is the contour pixels of the ruler generally having a higher contrast than the surrounding pixels,the area of the ruler is always closed and the scale of the ruler always having a with the number of characters,and presents an end-to-end accurate positioning and identification of water level scale method.we perform image enhancements for nighttime badly illuminated images and super-resolution image reconstructions for low resolution remote water level images.After a large number of experiments,we have verified the improved detection accuracy and reliability of the algorithm.The main work of this paper is as follows:(1)We analyze and summarize some typical ideas and principles of the existing water level detections.And introduce the application of digital image processing in liquid level detection.The overall design of the water level detection algorithm based on digital image processing is introduced in detail,including image enhancement in nighttime or different conditions,scale recognition,water level output and so on;(2)We propose an end-to-end method for accurate positioning and identification of water level scale.The method first locates and intercepts the scale by adaptive threshold algorithm,morphological filter and edge detection algorithm,and then combine the histogram of Oriented Gradients and the Support Vector Machine method to identify the corresponding numeric characters.Compared with the traditional liquid level detection algorithm,this method can directly get the level of the scale display,and no longer need to manually identify.(3)Aiming at the night images with weak and uneven illumination,based on the traditional nighttime image enhancement algorithm,we propose a Retinex nighttime image enhancement algorithm based on guided filtering and global contrast enhancement.This method enhances the scale and background contrast while still maintaining the detail of the scale edge,which significantly improves the rate of water level recognition in the nighttime.(4)For the low resolution water level scale images obtained from the long shot,the details are not clear.We perform super resolution for low-quality scale images by using convolution neural network and deep recursive neural network in depth learning so that the reconstructed image can effectively represent the water level information and improve the rate of water level recognition under long-distance shooting conditions.