Abstract <p>Nondestructive testing of internal broken wires in wire ropes is crucial for maintenance. Existing methods focus on surface or shallow defects of exposed wires, neglecting detection at joints. The thick housing at wire rope joints makes internal broken wire detection difficult for common methods. Thus, this study uses CT technology to obtain wire rope cross sectional images and applies computer image processing for automatic defect detection. Traditional image processing methods struggle with complex cross sectional CT images, while deep learning based methods require extensive training data, which is hard to meet in practice. Therefore, a steel wire rope joint broken wire defect detection system based on template matching segmentation and gray level threshold judgment is proposed. First, the image is preprocessed, and a template is generated. Then, using constraint terms and the stochastic gradient descent(SGD), optimize the template to align with wire cross sections. Finally, an adaptive gray level threshold is calculated to judge defects at template circle locations. Experiments on generated wire rope joint CT image datasets show that the proposed system has stronger segmentation ability compared with general purpose segmentation models. And compared with object detection models, it can detect steel wire breaks more accurately and completely.</p>