During the towing process of light rescue wrecker

it is very likely that the trailer will leave the rescue wrecker due to excessive road bumps. In order to reduce security risk

a towing process security warning algorithm based on multiscale features fusion is studied. Firstly

the license plate is located acting as large-scale feature

Secondly

the license plate feature points are extracted based on the improved binary robust invariant scalable keypoints（ BRISK） algorithm

acting as the small-scale feature. Then

feature points between the current frame and the first one are matehed

based on the K nearest neighbor（ KNN） algorithm. Finally

the location information of multiscale features is fused to get the longitudinal amplitude of the sway of the tailed vehicle

thereby perform security warning. In addition

the proposed improved BRISK algorithm enhances the matching accuracy and image registration accuracy while ensuring real-time performance. The experimental results verify that the proposed algorithm has higher accuracy and better real-time performance.