I am attempting to run Stereo ORB-SLAM (EuRoC mode) with a custom dataset (https://www.eth3d.net/slam_datasets). However, I am not able to generate unified stereo parameters as seen in the EuRoC.yaml file in the Stereo Examples.
However, my problem is that for my dataset, I am given the intrinsics of both cameras (named calibration.txt and calibration2.txt), the extrinsics from the right camera to the left (named extrinsics_1_2.txt) and no distortion parameters (as the images pre-rectified). When I attempt to calculate the projection matrices for both cameras (using the left camera as a base) I am left with differing parameters for both cameras instead of unified values like in the EuRoC.yaml file.
My attempt to calculate the values are as follows, assuming no distortion:
P = K*[R|T]
where as [R_left|T_left] is an 3x3 identity matrix concatenated with a columns of zeros and [R_right|T_right] is the extrinsics given. My results using MATLAB are as follows:
Have I been performing the wrong calculation, or should I have included distortion parameters into my calculation? Any advice would be extremely useful, thanks.
Hi everyone,
I am attempting to run Stereo ORB-SLAM (EuRoC mode) with a custom dataset (https://www.eth3d.net/slam_datasets). However, I am not able to generate unified stereo parameters as seen in the EuRoC.yaml file in the Stereo Examples.
As shown below, the EuRoC.yaml parameters are:
Which is the same as the parameters found in the left and right projection matrices in the stereo rectification section:
However, my problem is that for my dataset, I am given the intrinsics of both cameras (named calibration.txt and calibration2.txt), the extrinsics from the right camera to the left (named extrinsics_1_2.txt) and no distortion parameters (as the images pre-rectified). When I attempt to calculate the projection matrices for both cameras (using the left camera as a base) I am left with differing parameters for both cameras instead of unified values like in the EuRoC.yaml file.
My attempt to calculate the values are as follows, assuming no distortion: P = K*[R|T] where as [R_left|T_left] is an 3x3 identity matrix concatenated with a columns of zeros and [R_right|T_right] is the extrinsics given. My results using MATLAB are as follows:
Have I been performing the wrong calculation, or should I have included distortion parameters into my calculation? Any advice would be extremely useful, thanks.
I've attached the following calibration files used for calculation: calibration.txt (left camera intrinsics) calibration2.txt (right camera intrinsics) extrinsics_1_2.txt (stereo extrinsics)