Traffic accident disability level assessment method based on hybrid prompt learning

To address the issues of current traffic accident disability rating systems, which heavily depend on the personal expertise and skill level of forensic experts, leading to low appraisal efficiency and high costs, a method based on blended prompt learning is proposed. First, combined cue learning methods and gated networks to build hybrid cue templates. Next, a pre-trained model with a Transformer architecture was introduced to learn the nonlinear mapping between rib fracture disability information and the corresponding disability level. Finally, transfer learning was applied using the pre-trained model with mixed prompt learning under small-sample data conditions to classify rib fracture disability levels. The results show that the accuracy of the proposed method is 86%, which is 29%, 30%, and 29% higher than baseline methods. The findings can effectively improve the efficiency of forensic appraisal and have practical application value.