Aocfs-ap clustering-based driving video key frame extraction method

Abstract

The application provides a driving video key frame extraction method based on AOCFS-AP clustering, input video frame data is vectorized through MobileNet to obtain vector data; the vector data is initialized by using ISCHMTP to obtain an initial solution matrix of the AOCFS algorithm, that is, the initial immersed object pose state of the AOCFS algorithm; the membership matrix and the membership matrix of the AP clustering algorithm are initialized; the fitness value of the immersed object is calculated, the fitness value is sorted to obtain a local optimal solution; the membership matrix and the membership matrix of the AP clustering algorithm are updated according to the local optimal solution; the motion state of the immersed object is updated according to the AOCFS update mechanism; if the obtained immersed object pose no longer updates or reaches the maximum iteration number, the loop is exited, the optimal clustering result is output according to the membership matrix and the membership matrix, otherwise the loop is repeated; the optimal clustering result is converted to a data frame sequence, and the obtained key frame data is output.

Images