To address the noise issue in experimentally acquired body pressure distribution data， this study proposes a convolutional autoencoder-based data reconstruction method to enhance data quality and usability. First， the body pressure distribution data is normalized. And Gaussian noise is added to construct the training set. A convolutional autoencoder model is designed and used for feature extraction and denoising. Subsequently， experimentally collected body pressure distribution data is utilized as the test set to evaluate the accuracy and stability of the reconstruction results. Experimental results demonstrate that the model achieves a mean relative error of 0.010 with a standard deviation of 0.018 across 98 test samples， indicating high accuracy and stability. Finally， the trained model is applied to process experimentally collected body pressure data， revealing the variation patterns of pressure distribution metrics with seat positions.