DIY Robotic arm calibration using OpenCV and chessboard pattern.

Рет қаралды 5,177

Жыл бұрын

Calibrating robotic arm is crucial for precise and reliable motion control. I wanted to find out if I will be able to calibrate kinematic parameters of my 3D printed robotic arm without using costly sensors, just utilizing OpenCV library and chessboard pattern. The main problem was, if it is possible to get reliable robot's position and orientation using camera attached to robot's end effector. For now I did it in my Unreal Engine simulator. Later I'll try this calibration procedure on the real robot. In this video I also cover such topics as : forward kinematics, Denavit-Hartenberg parameters and how to properly define them, how to compute Jacobian matrix, how to properly calibrate camera.
To support channel:
PayPal:www.paypal.com/donate/?hosted...
Patreon: / robo_age
Buy me a coffee: bmc.link/roboage21Y
Music:
"Dreams" from Bensound.com

Пікірлер: 23

@keepfighting8402 Жыл бұрын

Дякую, дуже вчасно підігнали відео!

@kot2905 Жыл бұрын

Good explanation, thank you!

@svitlana1367 Жыл бұрын

Thanks for the video

@vikavr5641 Жыл бұрын

Thanks!!! Good video!

@user-bs5us3xk3y Жыл бұрын

Good video

@likefoodua1088 7 ай бұрын

Дякую за відео

@billobama6018 11 ай бұрын

Hello, I'm a student who wants to use ROS2 in UE5 to test my image processing algorithm with the photorealistic images rendered by UE5. I found your great videos, and how can I find and install such a ROS2 connector for UE5? Thanks a lot!

@b-lifestyle7263 2 ай бұрын

Thanks for your video. I have some questions. If I have a point coordinate from camera coordinate. So, how robot can know to moving with its coordinate?

@roboage1027 2 ай бұрын

Thank you for your comment. You should know where your robot's camera is located. So you can create transform matrix from camera coords to robot coords. And then just use standard inverse kinematics to move the robot.

@b-lifestyle7263 2 ай бұрын

@@roboage1027 Thanks for your answer! But the important thing is the point coordinate that's camera coordinate,right? Hence, we need to transform the point coordinate to real word space coordinate first! And use it to create transform matrix from real world space to robot coords, right?

@roboage1027 2 ай бұрын

@@b-lifestyle7263 I think, you need only one transform - from camera coords to robot coords. Usually, robot's base is located in the world's origin. So, if you know transformation from camera frame to gripper frame, for example, you can use forward kinematics to calculate transformation from camera to gripper an fruther to base frame. And that will also be transformation from camera to world frame.

@b-lifestyle7263 2 ай бұрын

@@roboage1027 Thanks for your quick response! I follow the handeye calibration. And the relative equation AX=XB is popular. With A is matrix end-effector to base, B is chess to camera. Then, I assume I obtain the X matrix (including rotation & translation matrix) for transformation of camera coords to end-effector coords. For the test, I turn on the robot camera (rgb camera) and I move the mouse to 1 corner of chessboard and I get the x and y, so x and y is pixel, right? If it's pixel, then I product [x,y,z] with Mintrinsic to obtain [xc,yc,zc]. Finally, If I wanna get [x,y,z,rx,ry,rz] of end-effectors, I product the [xc,yc,zc] with X matrix, right?