紀要論文

Model Predictive Measurementを用いたステレオ360度カメラによる3次元人物追跡

Person 3D-Tracking with Stereo 360-Degree Camera using Model Predictive Measurement

jpn

24321192

https://doi.org/10.15002/00030621

法政大学大学院情報科学研究科

法政大学大学院紀要. 情報科学研究科編

19

1

6

2024-03-24

Version of Record

This paper propose a person 3D tracking method using stereo 360-degree cameras. In previous studies, it was difficult to estimate the 3D position of a person using only the image. This is mainly due to issues such as insufficient resolution and lens distortion in 360-degree images. I use Model Predictive Measurement(MPM), which enables meaningful measurement and recognition even in adverse conditions by utilizing knowledge of the object. The knowledge of the object was applied in the form of positional constraints on the person model. Two constraints were applied to the person model. First, the person must move on the ground. Second, the person must not move closer to the camera over time. The two constraints improved the accuracy of person 3D tracking. Person 3D tracking was possible for persons up to 20 meters away from the camera. This enabled the estimation of 3D coordinates at longer distances compared to previous methods. Calibration was performed to accurately determine the correspondence between the 3D coordinates of the person model and the images.

Article

AA12746425