Using Improved YOLOv5s for Defect Detection of Thermistor Wire Solder Joints Based on Infrared Thermography

Manual visual inspection for thermistor wire solder joints used in satellite circuits always results in inefficiency and low accuracy. A method based on improved YOLOv5s model is proposed to automatically inspect defections in thermistor wire solder joints through infrared images. In order to enhance feature extraction capability, ECA module is introduced to the backbone of YOLOv5s network. Image database is built through infrared image shooting platform for model training and testing. The experiment result indicates that precision, recall and mAP of improved model reach 64.1%, 92.0% and 88.4% respectively, which are all superior to original model. This work paves the way for automatically defect detection of thermistor wire solder joints based on infrared thermography.