An Automatic Detection and Recognition Method for Pointer-type Meters in Natural Gas Stations

Currently, pointer-type meters in natural gas stations are read manually, which causes a waste of human resources. Researchers have proposed lots of methods to realize automatic reading of such meters. However, these methods need high quality source image rigorously so that they cannot perform well in real environments. To solve this problem, an automatic detection and recognition method for pointer-type meters is proposed, which consists of R-FCNs (region-based fully convolutional networks). an improved local threshold segmentation method and a probability-circle method. The experimental results show that the method performs well when the source image is blurred, the illumination is complex or the camera's angle is tilted. It has good robustness and accuracy to detect and recognize the pointer-type meters automatically.