jedeschaud / ct_icp

CT-ICP: Continuous-Time LiDAR Odometry
MIT License
756 stars 130 forks source link

kitti timestamps #2

Closed dongfangzhou1108 closed 2 years ago

dongfangzhou1108 commented 3 years ago

great appreciate with opening your great work! first, can you provide dataset passcord to me, no one relpy in mail, second, And when I see kitti website, the raw point cloud data do not have time information except scan begin and end time, How do you get the time information?

thanks a lot best wishes!

jedeschaud commented 3 years ago

Hello, The password for the dataset folder is : npm3d CT-ICP does not need the real timestamp of points but only the relative timestamp inside one scan. For KITTI dataset, we compute the timestamp from the azimuth of points in a scan. Here the Python line to compute the timestamp : data[:,3] = float(i-FIRST_INDEX_FRAME)/10.0 + (-np.arctan2(data[:,1], data[:,0])/np.pi+1.0)*0.05

Best Regards

dongfangzhou1108 commented 3 years ago

Thank you very much!

dongfangzhou1108 commented 3 years ago

sorry,Can you provide the python code for me?thanks a lot!

jedeschaud commented 3 years ago

PLY_Scans_from_KITTI_raw_data.zip

dongfangzhou1108 commented 3 years ago

thanks for your code, but when i check the code, data[:,3] = float(i-FIRST_INDEX_FRAME)/10.0 + (-np.arctan2(data[:,1], data[:,0])/np.pi+1.0)*0.05, by experiment, i find the beginning of the scan is from back of the lidar, can you explain it to me? thanks a lot,

here is my experiment, this is a scan ,and the height(z) is the output of float(i-FIRST_INDEX_FRAME)/10.0 + (-np.arctan2(data[:,1], data[:,0])/np.pi+1.0)*5, (ten times of orgin, for viewing), by matlab image

dongfangzhou1108 commented 3 years ago

i have seen in ct_icp code, you use the max and min of time to compute, but i guess there have a problem?

thank u very much

pdell-kitware commented 3 years ago

thanks for your code, but when i check the code, data[:,3] = float(i-FIRST_INDEX_FRAME)/10.0 + (-np.arctan2(data[:,1], data[:,0])/np.pi+1.0)*0.05, by experiment, i find the beginning of the scan is from back of the lidar, can you explain it to me? thanks a lot,

here is my experiment, this is a scan ,and the height(z) is the output of float(i-FIRST_INDEX_FRAME)/10.0 + (-np.arctan2(data[:,1], data[:,0])/np.pi+1.0)*5, (ten times of orgin, for viewing), by matlab image

Hi,

This is not a problem, KITTI's Lidar frames start behind the car, and turn clockwise (when considering the direction of motion),