【作者】 陈红；

【导师】 唐洪武；

【作者基本信息】 河海大学 ， 测试计量技术及仪器， 2005， 硕士

【摘要】 本文是在国家自然基金项目“多沙河流实体模型表面流场、河势测试理论与分析技术研究(50479068)”的资助下完成。 对于实体模型,流速测量方法有多种,但多数方法不能实现多点测量,为此本文建立了一套实体模型表面流场测试系统,用于实体模型表面速度场的测量;同时,结合边缘检测、轮廓提取和神经网络技术提取河势数据。文中详细介绍了实体模型表面流场、河势测试系统的原理和实现方法。在模型河流表面散播示踪性能好的粒子,通过摄像头获取流场图象,将图像数据传输到计算机,经过图像预处理后,根据示踪粒子的浓度,分为PTV和PIV处理方法。PTV处理中,为了获得粒子像的代表点,本文提出了像缩小法、权值提取法和圆拟合法,再结合流体运动特征进行粒子匹配,从而获得流速信息。PIV处理时,利用Hartly变换进行初始互相关运算,并对相关系数进行曲面拟合,通过曲面最值求解获得最大相关系数,以提高速度测量精度。河势处理时,根据流场图像对比度的不同,通过对各种边缘检测算子的比较,选取处理效果较优的算子,并依据流体运动特性建立了三层感知器神经网络模型,获得河势数据。 本文采用Visual C++ 6.0语言编制了包含采集、处理和保存三个主要功能模块的测试系统软件,并在实体模型上得到应用,文中还对实验结果和误差来源进行了分析。更多还原

【Abstract】 This paper was supported by the National Natural Science Foundation of China. "The research of theory and technique about Surface Flow Field and River Regime in Sediment-laden River (50479068) "There are a lot of ways to measure velocity, but few can measure muili-point in physical model. As a result, This paper have set up a measuring system which can measure velocity field, meanwhile, applied edge detection technique, contour extraction Technique and neural network Technique, to get the data of river-regime. Paper introduced the detail of principle and realization ways about the system. Broadcasting seeding which can follow stream current and useing CCD cameras to record images of flow field, sending the images to computer, then classifying to PIV or PTV treatment by density of seedings in images. In PTV treatment, paper has founded image-reducing law, weighted law and circle fitting law for getting point of particle image, then applied hydrodynamic rule, we can get velocity of flow. In PIV treatment, paper has founded cross-correlation and surface fitting law, applying Hartly transformation in cross-correlation to get prime correlation coefficient, then doing surface fitting of correlation coefficient, we can get the most correlation coefficient derivation, this law can improve precision to subpixel. In river-regime treatment, it is first to compare effect of all edge-operator and choose edge operator with the best effect, then applied three-layer perceptron neural network in contour extracting to get data of river-regime.The software of system was compiled by visual C++6.0, including the capture, processing and save module. Paper has applied the system into physical model, and analyzed the result and error.更多还原