Traditional exercise rehabilitation mode information is closed, it is difficult for users to obtain exercise rehabilitation information, and the exercise rehabilitation data of users are not fully utilized. With the combination of big data technology and the Internet of Things and exercise rehabilitation, we also face 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 sports data processing and combine with the big data processing technology to explore the factors affecting exercise rehabilitation to improve the effectiveness 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 carries out experimental analysis to verify the effect of the method proposed. The research results show that the method constructed in this paper has a good physical exercise rehabilitation effect.