【作者】 贾谊；

【导师】 严波涛；

【作者基本信息】 上海体育学院 ， 运动人体科学， 2012， 博士

【摘要】 影像测量方法是运动生物力学研究中的常用测量方法。对人体动作规律的揭示主要是依靠影像测量方法的测量结果，而影像测量误差的存在会产生伪信息，歪曲混淆对人体动作现象的认识，甚至会导致出现错误信息和结论。为获得可靠的测量结果，必须研究影像测量过程中误差的大小、关系和性质，研究消除、抵偿和减弱测量误差的措施。在20世纪70-80年代期间，影像测量方法开始在体育科学研究中广泛使用，也是在此期间，出现了的大量关于影像测量方法测量误差的文献和研究成果，但这些研究结果大都只是零散的，着眼于影像测量过程中的某些环节出现的具体问题而进行的探讨，研究成果不具有系统性，研究内容也不全面。因此，我们有必要对影像测量方法的误差问题进行系统的分析，全面考察影响影像测量精度的误差因素、误差性质以及减小和消除误差的对策与方法，从而使研究者能够更加深入的了解影像测量方法的误差范围，避免误差的产生，达到正确、有效利用这一方法为体育科研工作服务的目的。研究通过设计不同的拍摄条件，对平面定机拍摄和立体定机拍摄方法的误差大小进行了实验测试，主要研究结果如下：（1）取景范围中间3/5的部分是畸变误差较小的区域，基本上不受镜头畸变的影响。平面定机拍摄时，拍摄距离和测量误差值之间没有明显的线性相关关系。（2）立体定机拍摄时，摄像机机位的放置对测量精度影响较大，以主光轴夹角90°最为适宜。（3）在标定空间范围内，内层空间精度最为稳定，而外层空间精度较差。在标定空间范围外，随着被测量框架远离标定中心，测量误差有逐渐增大的趋势。（4）从测量结果来看，随着标定框架使用时间和使用次数的增长，框架各标定点的坐标值会出现不同程度的偏移，三个轴方向的偏移量平均值不超过0.01m。（5）利用现有的图形图像处理技术可对平面定机拍摄中的镜头畸变误差以及透视误差进行快速修正。（6）在使用数字滤波法对原始数据进行平滑时，一定要根据拍摄频率来确定数字滤波的截断频率，否则可能会导致截断频率过低而滤掉有用信号或者过高而使得噪声进入的结果。（7）利用多因素方差分析方法的分析结果表明，多个解析员同时对同一动作视频进行解析时，测试者之间的差异有时足以达到掩盖真实动作差异的程度，从而使解析结果可靠性大大降低。主要结论：（1）畸变误差是平面定机影像测量方法的主要误差来源，镜头边缘处畸变量随镜头不同差异较大，在进行拍摄时，应尽量选择畸变量小，性能稳定的镜头从而避免和减小畸变误差对测量结果的影响。（2）在进行立体定机拍摄时，摄像机位置的改变对垂直于地面方向的测量精度影响不大，而对平行于摄影基线方向的测量精度影响较大。摄像机主光轴夹角为90°时，测量精度最高。（3）借助目前的图形图像处理技术和处理软件，可针对影像拍摄过程中的某些误差进行修正，经实验论证，可以提高测量精度。（4）解析人员对人体关节点的判读误差是影像测量过程中最主要的误差来源，由不同解析人员对关节点位置的判断差异，足以达到掩盖真实动作差异的程度，从而使解析结果可靠性大大降低。更多还原

【Abstract】 Image measurment and analysis is an usual measurement method inbiomechanics resarch. We can find the interpretation of human body’s movementrules primarily relying on the results of this kind of measurement. But imagemeasurment can also result false information and twist to the phenomenon of the bodymovements even lead to error messages and conclusion. To obtain reliablemeasurement results, we must study the size, relation and nature of the measurmenterrors, moreover, we must find some methods for errors reduction.In the1970s and1980s, the image measurement method has been widely used insports science research, meanwhile, a mount of literature and studys on the measuringerrors of this method has arised. But these findings were just fragmented and onlyconcentrate in some areas of specific problems without systematicness.Therefore, it is necessary to analysis the measurement errors in sports film andvideo, give a comprehensive inspection for the impact factors, the nature andreduction of errors. Thus, the researchers should carry on further studys on the size oferrors, avoid errors and use this method for scientific research serve correctly andeffectively for the purpose of the sports research.By constructing different capture conditions, thie study test the size of errors fortwo-dimentional and three-dimentional image measurment. The results are as follows:(1)In the mid of3/5of the view area, the distortion error of lens is small. In twodimentional analysis, there is no obvious linear relationship between camera distanceand measurement errors.(2)In three dimentional measurment, the precision is greatlyinfluenced by the camera position, the most suitable angle of cameras’ axis is about90°.(3)In the caliberation volum, the precision of inside layer is most stable, theprecision of outside layer is the lowest. Out of the caliberation volum, as the object isfar from the framework, the amout of error have a tendency to grow.(4)From themeasurements result, with the number and time of use, the coordinate values of theframwork would appeared some deviation. The average errors of all three axis are notmore than0.01m.(5) By the software photoshop and lensdoc plug-in, we can fix thedistortion errors and out-of-plane errors in time.(6) By using image measurmentmethod for action and technical quantizing analysis, we must determine the cut offfrequency of digital filtering according to the film frequency, otherwise, it is probablelead to a low cut off frequency and filter out some useful signals or high cut offfrequency and leave too much of noise.(7)The result of ANOVA analysis has showedthat if there are multiple digitizers to digitise a same movement file, the movementvariability cannot be determined reliably.The main conclusion of this text are as follows:(1)Distortion error is the mainerror sources in two dimentional image measurment method. The distortion error inthe edge of lens is depend on the camera type. Before a motion capture process, weshould choose a small distortion and stable performance of the lens in order to avoidthe distortion error which influence the measurement results.(2) In three-dimensional capturing, the location of camera can not affect the measurement accuracy whichperpendicular to the ground direction, and greatly influenced the accuracy of themeasurement accuracy which parallel to the baseline orientation of camera. Thehighest accuracy of measurement emerged when the camera optical axis angle of90°.(3)With the help of the graphics and image processing technology and processingsoftware, we can get some certain error correction. By the experiment test, it can beproved that this technique can improve the accuracy of measurement.(4) The digitizeerror is the major error source in the analysis. Because of the different personneljudgment on joint points location, the movement variability cannot be determinedreliably, so that the analytic results of reliability is greatly reduced.更多还原