Research on Image Recognition Technology of Prefabricated Building Components Based on Convolutional Neural Network

By comparing the application effects of traditional image feature extraction algorithms (such as SIFT and ORB) and deep learning models (such as D2Net and SparseNCNet) in this field, this study innovatively integrates traditional features and convolutional features to build a RE-CB model. Experiments show that the RE-CB model can improve the Precision index of image matching by about 1–2 percentage points compared with a single algorithm on the Hpatches and Brown datasets, which verifies the superiority of fusion features in image recognition of complex prefabricated building components, and provides a new idea and an effective solution for the development of image recognition technology in this field.