中文摘要：

开发了一种用于河流表面流速场测量的便携式大尺度粒子图像测速仪（large—scale particle image velocimeter，LSPIV）。该仪器基于一台以图像处理专用DSP为核心的智能相机，代替Pc高效地完成图像采集和处理的任务。首先利用近红外波段下水流示踪物（弱小目标）和水体光强反射率的差异，在单色CMOS传感器前加装红外滤镜抑制水面光学噪声，增强弱小目标与背景间的对比度，提高了后续运动矢量估计中相关曲面的峰值信噪比。其次，提出了一种基于时空联合滤波的时均流场重建方法，利用河流流场中的先验和冗余信息识别并修正错误矢量，改善了天然河流中由于示踪物密度过低或分布不均造成的表面流速估值过低问题。最后，通过静水和动水条件下的2组现场实测验证了近红外成像系统及图像处理算法的有效性。

英文摘要：

A portable large-scale particle image velocimeter （LSPIV） system has been developed for river surface ve- locity field measurement. Based on a smart camera that takes an image-processing-oriented DSP as the core, the instrument can handle the image acquisition and processing tasks with higher efficiency than a PC-based system. Firstly, according to the difference of light intensity reflectivity between flow tracers （ dim targets） and water body under near-infrared （NIR） band,an NIR filter is mounted in front of the monochrome CMOS sensor to suppress the river surface optical noise. As a result, the contrast between dim targets and background is enhanced, and the peak signal- to-noise rate （PSNR） on the correlation surface in subsequent motion vector estimation is improved. Secondly, the paper presents a time-averaged flow field reconstruction method based on a spatial-temporal filtering strategy. By means of identification and correction of error vectors with the prior and redundant information in river flows, this method can overcome the underestimation problem of surface velocities caused by the low density and ununiform distribution of tracers in natural rivers. Finally, the effectiveness of the NIR imaging system and image processing algorithm have been verified by two sets of field experiments under static and dynamic water conditions.