about the speed calculation

[wenyan96]

Hi! Thanks for your great work firstly.

I have 2 questions when calculating the speed of objects(please see the following code). why times 36 in the sp_est_norm? From m/s to km/h --> supposed to be 3.6 times？And the formula 24 in the paper, why not use the timestamps of consecutive frames as the delta_t to calculate the speed?

// // ***** get the ground truth object speed here ***** (use version 1 here)
cv::Mat sp_gt_v, sp_gt_v2;
sp_gt_v = H_p_c.rowRange(0,3).col(3) - (cv::Mat::eye(3,3,CV_32F)-H_p_c.rowRange(0,3).colRange(0,3))*ObjCentre3D_pre; // L_w_p.rowRange(0,3).col(3) or ObjCentre3D_pre
sp_gt_v2 = L_w_p.rowRange(0,3).col(3) - L_w_c.rowRange(0,3).col(3);
float sp_gt_norm = std::sqrt( sp_gt_v.at<float>(0)*sp_gt_v.at<float>(0) + sp_gt_v.at<float>(1)*sp_gt_v.at<float>(1) + sp_gt_v.at<float>(2)*sp_gt_v.at<float>(2) )*36;
// float sp_gt_norm2 = std::sqrt( sp_gt_v2.at<float>(0)*sp_gt_v2.at<float>(0) + sp_gt_v2.at<float>(1)*sp_gt_v2.at<float>(1) + sp_gt_v2.at<float>(2)*sp_gt_v2.at<float>(2) )*36;
mCurrentFrame.vObjSpeed_gt[i] = sp_gt_norm;

// // ***** calculate the estimated object speed *****
// estimated speed
cv::Mat sp_est_v;
// formula 24
sp_est_v = mCurrentFrame.vObjMod[i].rowRange(0,3).col(3) - (cv::Mat::eye(3,3,CV_32F)-mCurrentFrame.vObjMod[i].rowRange(0,3).colRange(0,3))*ObjCentre3D_pre;

float sp_est_norm = std::sqrt( sp_est_v.at<float>(0)*sp_est_v.at<float>(0) + sp_est_v.at<float>(1)*sp_est_v.at<float>(1) + sp_est_v.at<float>(2)*sp_est_v.at<float>(2) )*36;

cout << "estimated and ground truth object speed: " << sp_est_norm << "km/h " << sp_gt_norm << "km/h " << endl;

[wenyan96]

Hello, halajun
when reading the codes, I'm confused with some variables. Could you help me check if I have the correct understanding?
Frame::nSemPosition means the label in semantic instance mask of valid object classes
Frame::nModLabel means the tracking id of each object class but what's the difference between it and vObjLabel ?
when tracking the dynamic objects, the semantic label of the same car doesn't need to be the same in consecutive frames(eg. in Frame n, the label is 2, in the next Frame, the label is 3), the car can be tracked by optical flow(association between label2 and label3), and assign the same TrackId to the Obj in the consecutive frame. Is it correct? Thanks for your help!