A mobile interactive technology-based intelligent filling approach for educational game replies is researched and developed in order to improve students’ mobile interaction experiences in educational games. This method uses image recognition and location algorithm to process two-dimensional code image. On this basis, the model combines the STEM unite process (SUP) framework to build a SUP educational game model supported by mobile interaction technology. The performance of image recognition algorithm is analyzed experimentally. The experimental results showed that the average recognition time of the bilateral filtering algorithm was 31.2 ms. And the average recognition accuracy was 98.6%. The performance of image location algorithm is analyzed in the experiment. The final results showed that the positioning accuracy of the normalized correlation coefficient matching method under interference conditions was 93.5%. And the positioning time was 18 ms. Finally, the experiment analyzes the intervention of SUP educational game model on students’ academic performance. The result showed that using SUP educational game model can improve learning performance. To sum up, the SUP educational game model built by the research institute has certain help for learning.