Research on Water-Level Recognition Method Based on Image Processing and Convolutional Neural Networks

Water level dynamics in catchment-scale rivers is an important factor for surface water studies. Manual measurement highly accurate but inefficient. Using automatic sensors has disadvantages such as high cost and difficult maintenance. In this study, a recognition method based on digital image processing technology CNN proposed. For achieving batch segmentation of source images, the coordinates ruler region characters’ scale lines’ are obtained by using algorithms grayscale processing, edge detection, tilt correction Hough-transform morphological operations. The then used to identify value characters. Finally, calculated according mathematical relationship between number lines detected pixel traversal binarized This levels images collected Hulu watershed Qilian Mountains Northwest China. results show that accuracy compared with actual measured reached 94.6% improved nearly 24% template matching algorithm. With accuracy, low cost, easy deployment maintenance, can be applied monitoring mountainous rivers, providing effective tool hydrology research resources management.