CameraCaliPython

Introduction

The demo for single camera calibration by a 3d-rig picture. And it is compiled by Pycharm, use python3 and a library named numpy. There are several steps in this demo:

1. Choose a picture of 3D-rig online, as shown in Jietu20200301-091513.jpg Though we do not have a real rig, but this would not change the core method of camera calibration because it's okay to assume that the length and width of each small square is the unit length.
2. Mark some cross points by Labelme (or any other app with similar function) and record the pixel coordinate of them with the correspondings in world coordinate system. In this demo, the three intersections of planes are chosen as x,y and z axis of world coordinate system, which follows right-handed system. (as shown in Jietu20200301-091513.jpg) In this demo, 12 points are chosen. Generally speaking, though 6 points are enough theoretically, more points are good for the precesion of camera matrix estimation. In the image, three quadrangles are labeled by Labelme and only their vertex are useful,i.e., only their corrdinates would be recorded by labelme in image coordination system. The cooresponding value in world coordination system is recorded in pink in the image. The direction of the arrow indicates the order of coordinate recording.
3. Use the chosen points to work out M, K, R,t by methods on class.
4. To verify the calibration results, we bring in some points from the world coordination system and check whether they can be mapped to corresponding pixels by M.

Usage

To Use the demo, you can download from github to your own IDE, remember to change the sys.path.append and from..import.. row according to your own settings. Of course, the IDE is not necessary, you can alse run by command line if your python3 and numpy are ready.

CameraCaliPython

介绍

通过3D摄像机标定物的单张图像进行摄像机标定。 由Pycharm编译，使用python3和numpy的库。 此演示中有几个步骤：

1. 该示例选取网络上的一个立体标定物图片虽然没有真正的摄像机标定装置，但是可以假设每个小方块的长度和宽度都是单位长度。这与摄像机标定原理并不违背。
2. 用Labelme（或具有类似功能的任何其他应用程序）标记一些交叉点，并在世界坐标系中记录对应点的世界坐标。在此示例中，选择三个平面的交线作为世界坐标系的x，y和z轴，该坐标系遵循右手系。 （如 Jietu20200301-091513.jpg 中所示） 在此示例中，选择了12个点。尽管理论上6个点就足够了，但更多点有助于提高摄像机标定的精确度。 在图像中，三个四边形用Labelme标记，只有它们的顶点才有用，因为labelme只会记录顶点的像素坐标。世界系统中的对应值以粉红色记录在图像中。箭头的方向表示坐标记录的顺序。
3. 使用选定的点通过课上的方法算出M，K，R，t。
4. 为验证校准结果，我们从世界坐标系中引入一些点，并检查它们是否可以通过M映射到相应的像素。

用法

要使用该演示，您可以从github下载到自己的IDE，请记住根据自己的设置更改 sys.path.append 和 from..import .. 行。 当然，IDE不是必需的，如果您已经安装python3和numpy，也可以通过命令行运行。