Algorithm 1 Predicting the classification line

Input: best classification line and observed centers of targets at time $k$; Output: predicted classification line and observed centers of targets at moment $k+1$;  1: Input the best classification line and observed centers of targets at time $k$ as initial  parameters, get the reference classification line and predicted targets positions $A_{k+1|k}$  and $B_{k+1|k}$ at moment $k+1$;  2: Calculate the intersection point $C_k$ and scaling factor ${\alpha }_s$ according to the straight line  equations;  3: Compute the predicted intersection point $C_{k+1}$ on the basis of scaling factor ${\alpha }_s$ and  predicted positions of targets $A_{k+1|k}$,$B_{k+1|k}$;  4: Calculate the normal vector of $l_{k+1|k}$ based on positions of targets at time $k$ and $k+1$;  5: Obtain the predicted classification line at moment $k+1$ with intersection point $C_{k+1}$ and  normal vector of $l_{k+1|k}$;  6: Calculate the observed centers of targets at time $k+1$ after classifying the sampling points  with the predicted classification line, which will be utilized to track targets based on EKF.