Non-contact motion measurement of mastication by integration of multiple sensors

下颌运动测量对临床颌骨功能诊断具有重要意义。下颌运动的三维测量对于诊断咀嚼功能尤其重要，因为咀嚼功能反映了下颌关节的状态，这些下颌关节通常从外观上看不出来。下颌运动测量的一种替代方法是测量嘴唇周围的面部运动。该运动明显反映了咀嚼相关运动。本研究开发了一套立体视觉面部运动测量系统，并提出了一种多传感器融合算法，以实现从外观上诊断牙齿疾病。1）多传感器数据融合提出了一种融合距离传感器数据和立体视觉数据的融合算法。该算法基于贝叶斯估计和卡尔曼滤波。2）嘴唇及邻近区域多点运动测量系统研制了嘴唇及邻近区域多点运动测量系统。在该系统中，15个标记贴在嘴唇周围的皮肤的三维坐标是由立体视觉跟踪。利用该系统，我们分析了下面部运动与颌骨、肌肉和咬合紊乱之间的关系。该方法将时间序列数据分解为主成分。一般来说，下面部运动可以用两个或三个分量来精确描述。3）实时测量系统的设计我们更新了我们第一年开发的图像采集系统。新的唇动测量系统可以记录10秒的图像序列，并实时分析数据点。在研究之初，我们计划使用DSP芯片，但由于CPU技术的进步，我们能够在没有DSP的情况下构建系统。

Motion measurement of lower jaw is important for clinical diagnosis of jaw function. Three dimensional measurement of lower jaw movement is especially important in order to diagnose mastication function which reflects jaw joint status ; these jaw joints are usually unseen from appearance. An altenative of the jaw motion measurement is the measurement of facial motion around lips. The motion obviously reflects mastication-related motion. In the research, we developed a stereo vision facial motion measurement system, and an integration algroithm of multiple sensors in order to diagnose dental disorder from appearance.1) Integration of multiple sensor dataAn integration algorithm has proposed to integrate range-sensor data and stereo vision data. The algorithm is based on Baysian estimation and Kalman filter.2) System for multi-point movement measurement of lips and adjacent regionA multi-point measurement system has developed. In the system, the 3D coordinates of 15 markers stuck to the skin around lips are tracked by a stereo vision. With the system, we analyzed the relationships beween lower-face motion and disorders of jaws, muscles, and occlusion. The method decomposes the time-series data into principal components. Generally, lower-face motion can be precisely described with two or three components.3) The design of realtime measurement systemWe renewed the image capturing system which we developed in the first year. The new lip-motion measurement system can record 10 second image sequences, and analyze the data points in realtime. In the beginning of the research, we planned to use a DSP chip, however, we were able to construct a system without DSP because of the progress of CPU technology.