In this experiment we use a stationery system with two identical cameras with parallel axes to track a moving object in three dimensional space. We capture a moving object in simultaneous video clips and extract only the moving object in each frame in the clips. We assume that the filters remove all the clutter in the subtracted images. Our tests indicate that it is the case for indoor imaging under average illumination conditions. Performance of the non-static filter for outdoor images was satisfactory when moving components of the background remains within a window of a limited size and brightness of the background remains fairly constant. Consider two frames, one from each camera, capturing the moving object at the same instant by a parallel axis camera system. The top most point of the moving object will appear in both images and more importantly, the pixels corresponding to the top most point of the object lie in the same pixel row in both frames. If the two images are free of background clutter, the pixel row in question is the top most pixel row of each image that contains an
object pixel. (It should be clear that the top most pixel rows of the images that contain an object pixel must have the same row index.) This simplicity of the epipolar geometry of the camera system with parallel axis allows effortless identification of the top most points of object in two images. Then (x,y,z) coordinate of the real world position [3] of top most point of the object is given by Eq. (1),(2) and (3).

In order to do this we use the top most, bottom most, left most, and right most point of the object. Very often block matching yields wrong matching pairs due to the existence of more than one point on the same horizontal line in the image with similar intensity properties. To avoid this inaccurate matching we match several points close to the center of the object and pick the most promising matching by checking of relative location accuracy. This technique is described in details in .

内容多 帮帮了 谢谢啦～～～～～～～～～～～～～～～～～～～～～～～～～～～～～～～～～·

In this experiment we use a stationery system with two identical cameras with parallel axes to track a moving object in three dimensional space.
这个实验，我们用一种用平行轴固定的两台同步摄像机定位系统来跟踪一个三维空间中的移动物体。
We capture a moving object in simultaneous video clips and extract only the moving object in each frame in the clips.
我们从同步拍摄中捕捉移动的物体，并且从每一帧图像中截取移动物体。
We assume that the filters remove all the clutter in the subtracted images.
我们假设用滤镜将这些截取出的图像中的所有多余混乱模糊的影像去除。
Our tests indicate that it is the case for indoor imaging under average illumination conditions.
我们可以看到在均衡的照明情况下，室内成像的问题。
Performance of the non-static filter for outdoor images was satisfactory when moving components of the background remains within a window of a limited size and brightness of the background remains fairly constant.
当在一个固定尺寸的窗口移动背景及清晰明亮的背景下移动，对于室外成像，动态滤镜的表现更让人满意。
Consider two frames, one from each camera, capturing the moving object at the same instant by a parallel axis camera system.
通过平衡轴摄像系统截取移动物体的同一瞬间的两帧图像，两台摄像机各取一帧。
The top most point of the moving object will appear in both images and more importantly, the pixels corresponding to the top most point of the object lie in the same pixel row in both frames.
两张图像中会显示出移动物体的顶点，更重要的是，两帧图像中物体顶点对应的像素点处于同一像素列。
If the two images are free of background clutter, the pixel row in question is the top most pixel row of each image that contains an object pixel.
如果两个图像中背景不混乱模糊，那这个像素列很有可能是每个图像物体像素的顶点像素列。
(It should be clear that the top most pixel rows of the images that contain an object pixel must have the same row index.)
像素图像的顶点像素列必须参数相同，图像才能清晰。
This simplicity of the epipolar geometry of the camera system with parallel axis allows effortless identification of the top most points of object in two images.
这个平衡轴摄像系统运用简易几何学很容易的说明了两图像中物体顶点像素问题。
Then (x,y,z) coordinate of the real world position [3] of top most point of the object is given by Eq. (1),(2) and (3).
然后，通过实验的步骤的1、2、3分别得出物体在顶点位置x、y、z坐标。
In order to do this we use the top most, bottom most, left most, and right most point of the object.
为了得出以上结果，我们分别测试了物体的顶点、底部、左端、右端的像素点。
Very often block matching yields wrong matching pairs due to the existence of more than one point on the same horizontal line in the image with similar intensity properties.
但由于同一亮度的图像中，同一水平线上存在多点，以至于造成部分结果不匹配。
To avoid this inaccurate matching we match several points close to the center of the object and pick the most promising matching by checking of relative location accuracy. This technique is described in details in .
为了避免这种匹配错误的现象发生，我们选择靠近物体中心的几点进行匹配，并且通过彼此位置的调整来选取最接近的一点。这个方法已详细说明。

其中可能有欠妥的地方，仅供参考。^o^||