Abstract: In the process of hydropower engineering construction, a large number of non-editable documents for hydraulic concrete materials are generated. Using manual interpretation methods to obtain texts is time-consuming, laborious, and the uncontrollable in accuracy, making it difficult to meet the demand for information management of material data. To address this issue, this paper proposes an intelligent interpretation method for non-editable texts of hydraulic concrete materials. Firstly, a text detection model called HC-PSENet based pixel level segmentation is constructed, which integrates the backbone network of PP-HGNet to achieve accurate detection of text lines. Furthermore, a professional corpus based on domain knowledge is created to obtain accurate character mapping. The text recognition model named HC-CRNN for hydraulic concrete materials is established using detection text boxes and professional corpus as inputs. The backbone network of ResNet and improved loss function C-CTC Loss are used to improve the accuracy of character classification. Finally, taking the self-made dataset as an example, the transfer learning strategy is introduced to train the model. The effectiveness and superiority of the proposed method are verified through ablation and comparative experiments. The results show that the proposed method has a harmonic mean of 0.985 for detecting text regions and the accuracy of text recognition reaches 90.62%. Its overall performance is superior to classical methods, aiming to provide new technical means for the automated reuse of non-editable resources in concrete materials.