search

hengguan / Hand-Eye-Calibration-Matlab

hand eye calibration include eye on hand and eye in hand
61 stars 23 forks source link

Detile Question #2

Open GuancongLuo opened 4 years ago

GuancongLuo commented 4 years ago

Hello, I am glad to see your open source contributions and "JIANSHU" sharing.

When I used your code to perform eye-to-hand experiments, I got a very large error (154Pixel Error). There are a few details and I hope you can talk to me.

  1. What material is your calibration plate? （ Papery? or Hard board?)

  2. How to deal with images, what criteria should qualified images meet?

  3. How do you adjust optional parameters when there are still large errors?

Hope you can reply to this issue in your busy schedule, thank you very much!

您好！很高兴开到您的开源贡献以及简书分享。 我在使用您的代码进行眼在手的实验时，得到了很大误差的结果（154Pixel Error）。有几个细节问题希望您能和我交流一下。

  1. 你的标定板是什么材料？纸质的？硬板的？
  2. 如何处理图像，合格的图像应该满足什么指标？
  3. 依然有很大误差时，您是如何调整可选参数的？

希望您能在百忙中回复该issue, 十分感谢！！

ChanJY0623 commented 3 years ago

Hello, I have the same problem as you.I calibrated it with my eyes and hands, using a 25mm, 7*10 checkerboard. At the same time, I observed that the translation vector (t) in the pose conversion matrix in the example images (./examples/ Calib1211) provided by the author should be in meters (m). After I modified my matrix based on this conjecture, there was still a large error (187.6 pixel Error).

您好，我和您有同样的问题。我使用眼在手标定，使用25mm，7*10棋盘。 同时我观察到，作者提供的示例图片（./examples/calib1211）中，其位姿转换矩阵中的平移向量（t）单位应该为米（m）。 我根据这个猜想修改了我的矩阵后，仍然有很大的误差（187.6 pixel Error）。

Twilight89 commented 3 years ago

@GuancongLuo @ChanJY0623 Hi, did you guys solve this problem? When I use this code(eye-to-hand), I got an error 55.5 pixel Error, which is very large. I don't know what problem happened, can you help me ?

您好，请问你们解决这个问题了吗，当我使用这个代码时（眼在手外），我得到的误差是55.5 pixel Error（已经将平移向量转换为m） 我不知道问题出在哪里，请问你们有类似的解决方法吗？ Thanks ^^

Jonashs94 commented 3 years ago

@Twilight89: Try to modify baseEst and inliers in the CalCamArmEIH.m script (trial and error). This worked for me, and I went from an error of 107 pixel to just 0.576 pixels.

By modifying baseEst only, my error went down to 1.621 pixel.

I realised that the camera calibration using the matlab calibrator app was OK in the beginning, so the problem must be with the optimization. Therefore, make sure your checkerboard images are OK by running the Camera Calibrator App in MATLAB and check the reprojection error. If everything is OK, then the problem is most likely with either the optimization or something could be messed up with your manipulator poses.

logeshg2 commented 7 months ago

@Jonashs94 How do I modify baseEst, when I change the inliers value from 80 to 5 the error becomes less than 10. Could you give me some examples? I think I am asking it too late.