Hi, I'm trying to solve the FOV issue as mentioned in #1034 (comment).
I've tested the usage of alphaScaling. The projection matrix can be calculated by cv::getOptimalNewCameraMatrix() correctly. But these data and images don't seem to work with rtabmap. I'm still looking for the reason, but the most likely reason is that using alphaScaling makes the image squeeze. That is to say, for the camera matrix calculated by cv::getOptimalNewCameraMatrix(), fx≠fy.
Here are the parameters I get.
left: fx=329.156006 fy=474.943451 cx=613.588454 cy=393.230591 baseline=0.075324
Here are the rectified left and right images (alpha=0.0). I put a coke in front of the camera, so you can see the squeeze.
Hi, I'm trying to solve the FOV issue as mentioned in #1034 (comment).
I've tested the usage of alphaScaling. The projection matrix can be calculated by cv::getOptimalNewCameraMatrix() correctly. But these data and images don't seem to work with rtabmap. I'm still looking for the reason, but the most likely reason is that using alphaScaling makes the image squeeze. That is to say, for the camera matrix calculated by cv::getOptimalNewCameraMatrix(), fx≠fy.
Here are the parameters I get.
left: fx=329.156006 fy=474.943451 cx=613.588454 cy=393.230591 baseline=0.075324
Here are the rectified left and right images (alpha=0.0). I put a coke in front of the camera, so you can see the squeeze.