Multi-scale instrument reading intelligent identification system for industrial site

Abstract

The invention discloses an industrial field-oriented multi-scale instrument reading intelligent identification system. The system comprises the following steps: acquiring instrument image data; an intelligent noise reduction and image enhancement processing module based on a multistage adaptive threshold segmentation algorithm and local area analysis is combined with a deep learning model (SSD) and a deep convolutional neural network (DCNN) model to obtain a pointer and dial plate identification result, and a coordinate system is constructed by taking the center of a circle of an identified dial plate as an original point; and calculating an instrument reading corresponding to the instrument image and using an obtained result to train a DQN reinforcement learning model so as to achieve an optimal identification effect. According to the method, manual reading is not needed, the instrument reading recognition accuracy is improved, the method is suitable for pointer reading recognition under the complex illumination condition of an industrial site, and the method has wide application prospects.

Images