A novel method of automatic reading for rotor water meter based on image processing. (14th June 2022)

• Record Type: Journal Article
• Authors: Wang, Juan; Li, Hongqing; Bai, Bing
• Is Part Of: International journal of simulation and process modelling. Volume 18:Number 1(2022)
• Page Start: 77
• Page End: 85
• Publication Date: 2022-06-14
• Subjects: rotor water meter -- automatic reading -- deep learning -- image processing -- multi-centroids method; Management -- Computer simulation -- Periodicals; Mathematical models -- Periodicals; Operations research -- Periodicals; Simulation methods -- Periodicals; 003.05
• Languages: English
• ISSNs: 1740-2123

Abstract

This paper investigates the issue of the reading of rotor water meters and presents a novel method of automatic reading based on image processing. In order to handle the impact of complex environment, we use object detection neural network to detect the bounding boxes of sub-dials on the water meter. Based on the standard spatial layout of sub-dials, the pose of water meter is corrected by perspective transformation. The regions of pointers are segmented from sub-dials by semantic segmentation. According to the segmented region, a multi-centroids method is proposed, through which the angle of the pointer area can be accurately obtained. The proposed method of automatic reading has better robustness and the obtained readings are more accurate. Simulation study is conducted to verify the effectiveness of the proposed method.