In order to conduct surface monitoring of the three-dimensional spine morphology of the human body in daily life, a spine morphology measuring method using "single camera, multi-view" to construct stereo vision is proposed. The images of the back of the human body with landmarks of spinous process are captured from multiple angles by moving a single camera, and based on the "Zhang Zhengyou calibration method" and the triangulation principle of binocular stereo vision, the spatial conversion matrices corresponding to each other between all images and the 3D coordinates of the landmarks are calculated. Then the spine evaluation angle used to evaluate the spine morphology is further calculated. The tests' results showed that the spine evaluation angle error of this method is within ±3°, and the correlation between the results and the X-ray film Cobb angles is 0.871. The visual detection algorithm used in this paper is non-radioactive, and because only one camera is used in the measurement process and there is no need to pre-set the camera's shooting pose, the operation is simple. The research results of this article can be used in a mobile phone-based intelligent detection system, which will be suitable for the group survey of scoliosis in communities, schools, families and other occasions, as well as for the long-term follow-up of confirmed patients. This will provide a reference for doctors to diagnose the condition, predict the development trend of the condition, and formulate treatment plans.
为了便于对人体的三维脊柱形态进行日常体表监测，本文提出一种利用“单相机，多视角”的方式构造立体视觉的脊柱形态测量方法。该方法通过移动单个相机使其从多个角度采集带有棘突标志点的人体后背图像，并基于“张正友标定法”和双目立体视觉的三角测距原理，求解出所有图像间两两对应的空间转换矩阵和标志点的三维坐标，由此推算出用于评估脊柱形态的脊柱评估角。试验结果表明，本文方法所得脊柱评估角误差在±3° 之内，结果与 X 线片测量法所得的柯布（Cobb）角相关性达 0.871。本文采用视觉检测算法，无放射性，且无需预先设定相机的拍摄位姿，操作简单。本文的研究成果可用于开发基于手机平台的智能监测系统，将适用于社区、学校、家庭等多种场合下的脊柱侧凸普查和患者的长期随访，为医生诊断病情、预测病情发展趋势、制定治疗方案提供参考。.
Keywords: Cobb angle; spine morphology; stereo vision; surface monitoring.