Algorithm 2. Steel wire rope surface defect detection method based on segmentation template and spatiotemporal gray sample set.

Inputs: Sequence images, time-domain frames $P$ Output: Sequence Images $D\left(x,y\right)s$ containing defects 1: get ROI mask with Formula (1) and Formula (2) 2: $t=0$ 3: while sequence images are not empty 4:  get ROI of the real-time image with ROI mask 5:   get $N\left(x,y\right)$ with RSCM (algorithm 1) 6:   $t=t+1$ 7:   if $t\le P$ 8:      use queues save $S_{off}\left(x,y\right),G_{off}\left(x,y\right) ,S_{on}\left(x,y\right),G_{on}\left(x,y\right)$ 9:   else 10:      get $\mathrm{gray}AVG$ of $S_{off}\left(x,y\right){{\displaystyle \cup }}^{}{G}_{off}\left(x,y\right) \mathrm{or} S_{on}\left(x,y\right){{\displaystyle \cup }}^{}{G}_{on}\left(x,y\right)$ 11:      if the gradient between $AVG$ and the current pixel $\le T_0$ 12:       update $S_{off}\left(x,y\right){{\displaystyle \cup }}^{}{G}_{off}\left(x,y\right) \mathrm{or} S_{on}\left(x,y\right){{\displaystyle \cup }}^{}{G}_{on}\left(x,y\right)$ by rule of         FIFO (first in first out) 13:     end if 14:     get $Q\left(x,y\right)$ with Formula (12) 15:     if $Q\left(x,y\right)\le T_1$ and $Q\left(x,y\right)\le T_2$ 16:      $D\left(x,y\right)=255$ 17:    else 18:      $D\left(x,y\right)=0$ 19:     end if 20:    open operation of the morphological method for $D\left(x,y\right)$ 21:    analyse $D\left(x,y\right)$ connected domain and remove small connected regions 22: end while 23: return $D\left(x,y\right)s$