Abstract: Pitting corroion is a hidden and highly destructive form of material damage, and conventional detection methods struggle to monitor its progression in real time. To solve this problem, a evaluation method for pitting severity based on acoustic emission (AE) image recognition was proposed in this paper. AE technology was employed to collect acoustic signals during the pitting corrosion process of Q235 steel. The K-means++ clustering algorithm was used to cluster the acoustic source signals, dividing the pitting corrosion process into three stages based on the composition of different signal sources. The acoustic source signals were converted into time-frequency diagrams, which were then used to train a convolutional neural network (CNN) model. The model was utilized to identify acoustic sources from the image signals and determine the corresponding stage of pitting corrosion. Finally, synchronized electrochemical data were analyzed to validate the accuracy of the evaluation method. The results demonstrate that the proposed method accurately classified the stages of the pitting corrosion process and could identify the acoustic source signals from Q235 steel pitting corrosion process with a recognition accuracy of 98%. This enabled the assessment of pitting severity based on the identified corrosion stage.