Development of a Field-Portable Omnidirectional Video System to Quantify Three-Dimensional Motion

为了研究自然环境下的运动，特别是户外运动，研制了一种电池供电的防风雨双摄像头视频系统和计算机分析系统。在数据采集系统中，每个摄像机都由人工操作人员定位和聚焦。相机将能够平移和变焦，以便可以很容易地跟踪拍摄对象。其他设计用于研究生物运动的系统要求摄像机的方向是固定的，但在这里，由两个光学轴编码器（分辨率0.036度）和变焦因子（由线性轴编码器测量）测量的平移和倾斜角度以视频帧速率（30赫兹）存储在条形码图像上。通过锁住摄像机并将相同的SMPTE时间码与每个视频信号混合，两个视频图像在时间上同步。图像处理系统由一个RISC工作站组成，该工作站运行软件来解码相机的平移和倾斜角度以及时间代码。此外，该软件还可以计算每个视频帧中多个物体质心的像素坐标。目标识别软件是由SVS公司的杰克·桑德斯-里德博士设计的，最初用于武器测试靶场。它具有像素处理、背景杂波抑制、目标检测和描述以及帧到帧跟踪等模块。相机角度和两个相机的物体质心的像素坐标用于计算物体在三维空间中的位置作为时间的函数。从这些基本信息中，可以计算出其他重要的性能特征，如速度、加速度、转弯半径和不同物体之间的距离。新的图像分析技术将有助于自动处理视频图像，如果手工完成是一个时间密集的活动。与其他15位科学家的一系列合作实验旨在测试该系统在野外条件下研究动物运动，动物行为和保护生物学方面的实用性和多功能性。这种仪器极大地提高了我们量化户外运动和行为的能力。它将提供数据收集的灵活性和图像分析的能力，这是任何商业运动分析系统都无法比拟的。

To study motion in natural settings, especially out of doors, a battery-powered weather proof two-camera video system and computer analysis system are developed. In the data collection system each camera is oriented and focused by a human operator. The cameras will be able to pan and zoom so that a subject can be easily followed. Other systems designed to study biological motion require that the orientation of the cameras be fixed, but here the pan and tilt angles measured by two optical shaft encoders (resolution 0.036 degrees) and zoom factor (measured by a linear shaft encoder) are stored at video frame rates (30 Hz) on the image in bar code. The two video images are synchronized in time by genlocking the cameras and mixing the same SMPTE time code with each video signal. The image processing system is consists of a RISC workstation that runs software to decode the pan and tilt angles of the cameras and the time code. In addition, the software can compute the pixel coordinates of the center of mass of multiple objects in each video frame. The object recognition software was designed by Dr. Jack Sanders-Reed of SVS Inc. and was originally used at weapons test ranges. It features modules for pixel processing, background clutter suppression, object detect and description, and frame to frame tracking. The camera angles and pixel coordinates of the center of mass of the objects from both cameras are used to compute the position of the objects in three-dimensional space as a function of time. From this basic information other important performance characteristics such as velocity, acceleration and turning radius and the distance between different objects can be computed. The new image analysis techniques will help automate the processing of video images, which if done manually is a time intensive activity. A series of collaborative experiments with 15 other scientists is designed to test the usefulness and versatility of the system for the study of a nimal locomotion, animal behavior and conservation biology under field conditions. This instrumentation greatly increases our ability to quantify motion and behavior outdoors. It will provide flexibility in data gathering and power for image analysis that no commercial motion analysis system can match.