Development of an engineering drawing detection and extraction algorithm for quality inspection using deep neural networks

Abstract

The manufacturing industry frequently produces multiple parts of the same type in a single batch process. Ensuring the quality of each part within specified tolerances is essential to maintain interchangeability in the assembly process. An integral aspect of quality inspection involves accurately interpreting engineering drawings to create inspection sheets. This study developed a deep neural network model to detect and recognize dimensions in engineering drawings for generating inspection sheets. The primary stages of the model include engineering drawing view detection, dimension detection, character recognition, information block processing, and output generation. The model was validated through two evaluation methods, k-fold cross-validation and testing on 10 real-world sample drawings, achieving a recall of 85.2%, precision of 88.7%, and an F-1 score of 86.9%. This proposed model can be implemented to reduce the time required for the quality inspection setup process, enhancing efficiency, and minimizing errors.