This paper aims to address the extraction and processing issues of multi-source heterogeneous detection data in smart metrology laboratories. It starts by analyzing the types of metrological detection data and the problems faced by the extraction of heterogeneous class information, proposing a detection data extraction technique based on image processing. Subsequently, an intelligent extraction technique for detection data based on differentiable binarization network (BDNet) and convolutional recurrent neural network (CRNN) is designed, achieving the automatic detection, recognition, and extraction of multi-source heterogeneous data. On this basis, a multi-source heterogeneous detection data intelligent extraction device was developed and verified. The results show that the device can effectively extract key information such as detection data from paper reports or forms, with high accuracy and rapid response speed, providing strong support for data management and analysis in smart metrology laboratories. This research is of significant importance for promoting the construction of smart metrology laboratories and the application of experimental detection data.