camera_2d_lidar_calibration
camera_2d_lidar_calibration copied to clipboard
ehong-tl
Why is there a negative value
Hi, Thank you for your work. When I add the following lines in the reprojection.py file "for i in range(len(img_points)): try: cv2.circle(img, (np.int32(round(img_points[i][0])),np.int32(round(img_points[i][1]))), laser_point_radius, (0,255,0), 1) print((np.int32(round(img_points[i][0]))),np.int32(round(img_points[i][1]))))"
The following values are printed to the terminal:
" ... (-401, 462) (-388, 462) (-374, 463) (-359, 463) (-344, 463) (-327, 464) (-310, 465) (-291, 465) (-272, 466) (-251, 466) (-230, 467) (-207, 468) (-183, 468) (-159, 469) (-132, 470) (-105, 470) (-76, 471) (-45, 472) (-13, 473) (20, 474) (55, 475) (92, 476) (130, 477) (170, 478) (213, 479) (257, 480) (303, 481) (351, 482) (401, 483) (454, 485) (509, 486) (567, 487) (628, 489) (691, 490) (757, 492) (826, 493) (898, 495) (974, 497) (1052, 498) (1134, 500) (1220, 502) (1310, 504) (1403, 506) (1501, 508) (1604, 510) (1711, 513) (1822, 515) (1939, 518) (2061, 520) (2188, 523) (2321, 526) (2460, 529) (2605, 532) (2757, 535) (2916, 538) ..." My question is, although the pixel width of the camera is 1280, why do I get values greater than 1280 or why are there negative values? Thank you in advance for your reply!
Hi @gokcesena
Values beyond the image boundaries (negative value or pixel value beyond image width in your current case) means that the 3D points projected outside of the image boundaries. It is normal to have this during reprojection. If you frequently encounter them, please try to refine your calibration.
Thank you for your quick reply @ehong-tl, the rmse value is high when I take samples frequently. I did it with 13 samples for calibration, but I will do what you said and take more samples. thanks
I'm wondering: Can I only publish point clouds that fall in front of the camera? I think this line: "objPoints = objPoints[Z > 0]" should give the pointcloud data falling in front of the camera, can you direct me?
Yes, these 2 lines of code will remove points that are falling behind the camera.
Z = get_z(q, objPoints, K)
objPoints = objPoints[Z > 0]