Chessboard mounting
AwsKhalil opened this issue · 3 comments
Thank you for your great work.
for chessboard detection from point cloud, does the chessboard need to be mounted on a stand or can a person hold it ? would that affect the detection of the chessboard ?
It is beneficial to use a stand. The reasons for that:
- The reflected points from the person may cause wrong segmentation results: the points from the person and the chessboard may be clustered to one segment.
- If the chessboard is not standstill in the hand of the person, the camera and the lidar should record the chessboard in the same time (synchronised).
Thank you. I figured that your first point is my problem. I couldn't perform the measurement again so here is how I solved it ( for anyone facing similar problem in the future) :
- run the 3 scripts normally (img_corner_est, pcd_corner_est, LM_opt)
- based on the results you get after running LM_opt script, find the odd images which show that there is a corner between the person and the chessboard and filter them out along with their point cloud from the original image,pointcloud pairs you had in the folders img and pcd
- rename the images and the point clouds and run the 3 scripts again.
Note: Make sure that the intrinsic parameters ('instrinsic_para' in the config file) are accurate.
I attached an example of an odd image
@AwsKhalil @ ValerZoldhegyi Could you tell me your enviroment of running ILCC. I cann't run the pcd_corners_est.detect_pcd_corners(). And I cann't solve it. Thank you very much!