受蜻蜓视觉机制启发的观测目标位移测量方法及应用

In recent years, river surface optical imaging velocimetry method, which is based on the conventional machine vision measurement, has been developed as a feasible means for non-contact river flow measurement during extreme flow conditions (for example, the flow velocity of the flood peak is more than 5 m/s). However, due to the complex and variable physical and chemical properties of flow tracers (objects or patterns floating at the water free surface), the various light disturbances (such as river surface reflected light, underwater scattered light, and atmosphere scattered light), as well as the rapid changes of the flow tracers' non-rigid states (such as scale, rotate, and occlusion), this kind of method is difficult to measure the displacement parameters of the observed targets (i.e., flow tracers) precisely. Then the reconstruction of flow field may be inaccurate, and finally the river discharge measurement would be failed. To solve these problems, this project will explore a novel observed targets displacement measurement method by imitating the spectrum-polarization imaging and the virtual small-eye-cluster image information processing mechanisms in a dragonfly vision. It can transform the spectrum-polarization sensitivities as well as the advantages of tracking water surface targets of compound eyes in the dragonfly, so it may break the bottlenecks of the traditional methods which cannot reconstruct the flow field and measure the river discharge steadily and reliably.

近年来国外发展起来的河流水面光学成像测速方法，是针对高流速的洪峰（大于5m/s）一类极端水流条件下进行非接触式水面流速测量的一种可行手段。但目前这种基于常规机器视觉测量的方法，对于水流示踪物（漂浮物或天然示踪模式）理化属性复杂多变，水面反射光、水下散射光及天空散射光等光线扰动，示踪物的尺度、旋转与遮挡等非刚体状态急速变化等情况，测量观测目标（水流示踪物）的位移参数比较困难，极易导致河面流场重建错误，最终造成流量测验失败。本课题试图提出仿蜻蜓视觉机制的光谱-偏振成像与虚拟"小眼群"视像信息处理联合的目标位移测量方法，探索转化蜻蜓复眼的光谱-偏振敏感性和追踪水面目标的优势，突破常规方法难以稳定、可靠地连续进行河面流场重建及流速测量的瓶颈。

提出了仿蜻蜓视觉机制的光谱-偏振成像与虚拟“小眼群”视像信息处理联合的目标位移测量方法，转化了蜻蜓复眼的光谱-偏振敏感性和追踪水面目标的优势，突破了常规方法难以稳定、可靠地连续进行河面流场重建及流速测量的瓶颈。通过仿蜻蜓大场景-小场景（LF-SF）系统的信息处理过程，对大视场下只占少量像素区域、且光学特性复杂的一类弱小水流示踪物实现了可靠检测及连续位移矢量提取。设计了虚拟“小眼群”，以读取场景光谱-偏振信息作为小眼群视像，通过模拟LF-SF系统中髓质细胞的基于脉冲序列做出判断的机制，设计了目标位移量测器。针对全光谱波段下水面成像噪声过大的缺点，通过实测数据分析确定了最佳成像波段，研制了0°、45°及90°方向光谱-偏振成像实验装置，解决了光谱-偏振成像光学结构设计及空间配准等关键技术。本项研究经现场技术验证和优化，形成了新型的河流水面光谱-偏振成像测速系统工作模式。为我国河流流量测验增添新型的仪器型谱提供了理论依据和技术方法上的支撑，使我国在极端水流条件下河流成像测速方面的理论成果达到或超过了国际先进水平，同时推进了机器视觉测量应用基础理论研究的发展。