Electrical Equipment Recognition System Based On Image Processing Technology

At present,the inspection of power plant equipment mainly relies on manual inspection.The inspection method has the problems of poor working environment,high labor intensity and high repeatability.The inspection robot can reduce the risk of inspection work in high-risk environments and improve the quality and efficiency of the inspection.Each electrical device of the power plant has a unique nameplate text code.The digital camera device,which is equipped with the intelligent inspection robot,is used to collect the image of the electrical device,and then the image processing technology is used to recognize the code on the nameplate of the electrical device to assist the robot in identifying the electrical device.In order to identify the electrical equipment of power plants,this paper researches the electrical equipment identification system based on image processing technology.The system consists of two parts: electrical equipment image detection algorithm and electrical equipment image recognition algorithm.The main research contents and research results of this paper are as follows:(1)A method for detecting the nameplate image text of electrical equipment based on the improved EAST algorithm is proposed.The ResNet network is used for deep feature extraction to suppress background interference.The experimental results show that the accuracy of the improved EAST algorithm is 6.1% higher than the original algorithm,which indicates that the method has better environmental adaptability.(2)Aiming at the problem that the difference in nameplate texts of different scale images greatly affects the accuracy of the nameplate text detection,a combination of the multi-scale training is proposed to realize the same algorithm to detect the nameplate text of different scale images;Aimingat the problem of text size imbalance in the same image,it is proposed to use balanced weight strategy to improve the accuracy of algorithm for detecting small texts.Experiments verify that the combination of multi-scale training and balanced weight strategy can effectively improve the accuracy of multi-scale image and small text detection.(3)An image recognition algorithm for electrical equipment nameplate based on STN+ResNet+BiLSTM+CTC is proposed.Firstly,the STN is used to correct the irregular text to the horizontal text to solve the problem that the irregular text is difficult to identify.Then,ResNet is used to extract the deep features,suppress the background interference,and then extract the context information of the text sequence through BiLSTM to avoid misidentification of non-text area into text,and finally use the CTC loss function to filter out the repeated characters,to solve the problem of no characters in some positions.The experimental results show that the proposed algorithm has higher recognition accuracy and faster computation speed.(4)The electrical equipment identification system software is designed and implemented.After the electrical equipment nameplate text detection and recognition algorithm is integrated,the software is written to identify the electrical equipment,and the effectiveness of the electrical equipment identification system software and algorithm is verified through experiments.