The goal of this study is to develop and validate a novel semi-automatic detection method (SADM) under nasopharyngoscopy based on the image processing technique, which can assist the evaluation of the velar closure ratio (VCR). After the development of the SADM, 72 patients were enrolled. The reliability of SADM was evaluated by repeated measurements. Velar closure ratio are given by conventional nasopharyngoscopy method and SADM were compared. Velar closure ratios given by SADM were further translated into a trichotomous classification for velopharyngeal function diagnosis, that is, velopharyngeal closure (VPC), marginal VPC (MVPC), and velopharyngeal incomplete. The 2 VCR-thresholds used for the translation were explored and validated. As results shown, SADM was proved to be reliable with an intraclass correlation coefficient of 0.996 (95% confidence interval: 0.993–0.997, P < 0.001). Intraclass correlation coefficient between conventional nasopharyngoscopy method and SADM was 0.954 (95% confidence interval: 0.927– 0.971, P < 0.001). Velar closure ratio-thresholds were set at 0.82 and 0.92 according to the ROC curve. Diagnostic sensitivity and specificity for velopharyngeal incomplete were 1.00 and 1.00. MVPC had 0.58 sensitivity and 0.96 specificities while VPC had 0.92 and 0.79, respectively. No statistically significant difference was found between the diagnosis of SADM and speech pathologists (P > 0.1). In conclusion, this study successfully developed an accurate and reliable semi-automatic method to evaluate VCR, which could help improve the efficacy of VCR evaluation and velopharyngeal function diagnosis.