Accurate estimation of the state of health (SOH) of lithium-ion batteries is crucial for optimizing battery design. However, accurate SOH estimation remains challenging due to the complex degradation mechanism within the battery. Therefore, a SOH estimation method based on charging voltage curve and physics-informed neural network (PINN) is proposed. Firstly, Spearman correlation analysis is used to extract battery aging characteristics from the constant current segment of the charging voltage curve and establish a partial differential equation model for battery SOH degradation. Secondly, using a neural network with added physics-informed constraints to approximate the implicit model. Then, the weighted average and convergence acceleration techniques of the weighted mean of vectors (INFO) algorithm are utilized to optimize the PINN hyperparameters and improve the estimation accuracy of the method. Finally, this method is used for SOH estimation on three publicly available datasets: MIT, CALCE, and NASA. The results show that the average RMSE of the proposed method on the MIT test set with changes in charging strategy is 0.271 6%. Compared with long short-term memory network (LSTM), convolutional neural network (CNN) and baseline neural network (BNN) methods, the error is reduced by 80.74%, 57.48% and 74.73%, respectively. The estimation accuracy on both the CALCE and NASA test sets is above 97%. This proves that the method has high estimation accuracy and good robustness to changes in electrode materials and experimental conditions.