Automatic Digital Recognition of Multiple Electricity Dashboards

摘要

Intelligent detection of electric transmissions and distribution equipment is impeded by factors such as the potential diversity of displays, the vulnerability of meter digits reading, and the slow speed of training and predicting. To deal with the problem, this paper aims to use the technology of pattern recognition to automatically recognize various kinds of digital displays by analyzing images collected by a CCD camera. To be more specific, firstly, we design a multi-layer convolutional module to classify different types of dashboards. Secondly, to deal with different types of dashboards, we presents an improved Adaboost algorithm in the training process that can quickly identify and cut out the display area of the meters. Lastly, we implement several digital image processing algorithms to recognize the dashboard readings accurately. The experimental results show that our method performs well in improving recognition precision.