3.1 A. Monocular, Close Stereo and Far Stereo Keypoints 单目、近处双目和远处双目特征点

3.1.1 paragraph

ORB-SLAM2 as a feature-based method preprocess the input to extract features at salient keypoint locations, as shown in Fig. 2b.

ORB-SLAM2作为一种基于特征提取的方法，在一些关键的位置上的提取进行预处理，如图2b所示，

The input images are then discarded and all system operations are based on these features, so that the system is independent on the sensor being stereo or RGB-D.

系统的所有运行都是基于输入图像的特征展开，而不依赖于双目或者RGB-D的相机。

Our system handles monocular and stereo keypoints, which are further classified as close or far.

我们的系统处理单目或者双目的特征点，分成远处特征点和近处特征点两类。

3.1.2 paragraph

Stereo keypoints are defined by three coordinates $\mathrm{x_{s}} = (u_{L},v_{L},u_{R})$ , being $(u_{L},v_{L})$ the coordinates on the left image and $u_{R}$ the horizontal coordinate in the right image.

双目特征点通过三个坐标定义 $\mathrm{x_{s}} = (u_{L},v_{L},u_{R})$，当中，$(u_{L},v_{L})$ 这个左边图像的坐标，$u_{R}$ 是右图当中的水平坐标。

For stereo cameras, we extract ORB in both images and for every left ORB we search for a match in the right image.

对于双目相机而言，我们提取两幅图像当中的ORB特征，对于每个左边的ORB特征我们对其匹配到右边的图像中。

This can be done very efficiently assuming stereo rectified images, so that epipolar lines are horizontal.

这对于建设双目图像校正十分有效，因此极线是水平的。

We then generate the stereo keypoint with the coordinates of the left ORB and the horizontal coordinate of the right match, which is subpixel refined by patch correlation.

之后我们会在左边的图像产生双目的ORB特征点，和一条水平的线向右边的图像进行匹配，通过修补相关性来重新定义亚像素。

For RGB-D cameras, we extract ORB features on the image channel and, as proposed by Strasdat et al. [8],

对于RGB-D相机，正如Strasdat等人[8]所言，我们提取在图像通道上提取ORB特征点，。

we synthesize a right coordinate for each feature, using the associated depth value in the registered depth map channel, and the baseline between the structured light projector and the infrared camera,

我们将深度值和已经处理的深度地图，和基线在结构光投影器和红外相机进行匹配，对每一帧的图像与右边图像的坐标系进行融合，。

which for Kinect and Asus Xtion cameras we approximate to 8cm.

这是kinect和华硕 Xtion 精度大约是8cm。

3.1.3 paragraph

A stereo keypoint is classified as close if its associated depth is less than 40 times the stereo/RGB-D baseline, as suggested in [5], otherwise it is classified as far.

近双目特征点的定义是：匹配的深度值小于40倍双目或者RGB-D的基线，否则的话，是远特征点。

Close keypoints can be safely triangulated from one frame as depth is accurately estimated and provide scale, translation and rotation information.

近的特征点能够从一帧的深度值能够三角测量化，是精确的估计，并且能够提供尺度，平移和旋转的信息。

On the other hand far points provide accurate rotation information but weaker scale and translation information.

另外一方面，远的特征点，能够提供精确的旋转信息，但是很少的尺度和平移信息。

We triangulate far points when they are supported by multiple views.

当提供多视图的时候，我们才能三角化那些远的点。

3.1.4 paragraph

Monocular keypoints are defined by two coordinates $\mathrm{x_{m}} = (u_{L},v_{L},u_{R})$ on the left image and correspond to all those ORB for which a stereo match could not be found or that have an invalid depth value in the RGB-D case.

单目的特征点通过右边图像的两个坐标当中的 $\mathrm{x_{m}} = (u_{L},v_{L},u_{R})$ 定义，必须保证所有的ORB特征是一致的，否则双目特征点的提取将不能够完整，或者在RGB-D的情况下，有产生一个无效的深度值。

应该是左图像帧，可能是笔误

These points are only triangulated from multiple views and do not provide scale information, but contribute to the rotation and translation estimation.

这些点仅能够从多视图中三角测量化并且不能够提供尺度信息，但是可以提供旋转和平移的估计信息。