In recent years, there has been a movement toward recording and analyzing dental treatment itself as an object of observation. However, the accuracy of identifying the treatment site is not high enough considering the application to medical safety and other aspects. This study aimed to improve landmark recognition accuracy by identifying the dental arch. By identifying the dental arch, 1) teeth detected outside the standard dental arch can be determined as false positives, and 2) the location of undetected teeth on the dental arch can be estimated using standard crown width diameter data, which improve the accuracy of identifying the treatment site. Therefore, we divided the video of a patient undergoing treatment into frame-by-frame image data, and started by determining the dental arch in these images. The dental arch is generally said to be parabolic. The axis of the parabola was identified from the facial landmarks, the vertex was taken on this axis, the opening of the parabola through the vertex and the dentition was identified, and the variance of the opening was obtained. The vertex was moved, and when the variance of the coefficient of the quadratic function was minimized, that parabola was considered the dental arch in the screen. As a result, the depiction of the dental arch was confirmed visually. Although the positional information of the landmarks in the oral cavity and on the face alone did not lead to a complete depiction of the dental arch, the accuracy of the dental arch detection model can be improved by deductively providing the recognition model with anatomical knowledge as a parameter. In addition, with a sufficiently accurate dental arch, we can expect to validate detected landmarks and identify undetected intraoral landmarks. It is possible to identify parabolas based on the location information of landmarks in the oral cavity and on the face.