Automatic Reading Recognition Method for Nuclear Power Digital Detection Instruments Based on Computer Vision and Deep Learning

Abstract Digital detection instruments are crucial monitoring devices in nuclear power plants, and the accurate recognition of instrument readings is of paramount importance for ensuring the safety, reliability, and economic operation of nuclear facilities. To address the challenge of low recognition accuracy caused by complex environmental factors and random noise affecting digital instrument images, this paper proposes a computer vision and deep learning-based method for automatic reading recognition of nuclear array detection instruments. Firstly, the digital detection instrument image sample set undergoes image preprocessing to generate higher-quality images. Subsequently, the limited digital detection instrument images are fed into a Deep Convolutional Generative Adversarial Network (DCGAN) to generate a large-scale training sample set. Building upon this, the YOLOv5 detection algorithm is employed to train the network using the generated sample set. Ultimately, the trained network achieves automatic reading recognition based on input images. The proposed reading recognition method in this paper completes an end-to-end recognition process, simplifying the recognition steps and exhibiting high recognition accuracy and robustness, achieving an accuracy of up to 98.74%.