Automatic Identification Method of Pointer Meter under Complex Environment

Abstract

As for the industrial pointer meter's recognition reading in outdoor scenes is easily affected by interference from complex environments such as light, occlusion, and blurred images, this article combines the theory and algorithm of case segmentation with pointer type in complex environments In meter detection and recognition, an automatic image segmentation and reading system for meter images was designed based on pyqt5. In the method designed in this paper, the de-noised image is firstly pre-processed by Non-Local Dehazing, Criminisi, and Deep Denoiseing Super Resolution, and then the instance segmentation algorithm Mask RCNN is used to segment and locate the pointer, and finally use the angle method for reading operation. Experimental tests show that this method can effectively overcome the interference of the complex environment on the pointer meter recognition, reduce the reading error, and has strong robustness and adaptive ability.