With the combination of big data technology and the Internet of Things and exercise rehabilitation, we are also facing the problem of processing massive amounts of data. Whether these data can be used well has become the key to obtaining good business benefits and improving user experience. This paper applies the Internet of Things and big data technology to the physical exercise rehabilitation system to improve the sports data processing and combine with the big data processing technology to explore the factors affecting the effect of exercise rehabilitation to improve the effect of physical exercise rehabilitation. Moreover, this paper proposes a posture recognition algorithm based on human vision, and defines the human posture described by the feature, which improves the classifier's ability to classify different posture data. Finally, this paper combines experimental analysis to verify the effect of the method in this paper. From the research results, it can be seen that the method constructed in this paper has a good rehabilitation effect of physical exercise.