The present invention relates to the technical field of child sitting posture correction. Provided are a child sitting posture detection method and system based on child facial recognition, by means of which method and system the age of a child is automatically recognized, and the sitting posture of the child is detected in real time and intelligently supervised according to the difference in the age of children. By means of the present invention, human body skeleton relationship information can be obtained by only performing a simple comprehensive mathematical operation on skeleton position information of several key parts of the human body, such as the eyes, shoulders, nose, legs, knees and feet; and the human body skeleton relationship information is then compared with a corresponding set threshold, such that the sitting posture state of the human body can be determined. In comparison with existing methods that require the use of high-order feature vectors, integral calculation, etc, only key data needs to be measured and calculated, without the need to perform separate model training on a sitting posture, such that the time and precision required for sitting posture detection can be greatly reduced.