Aiming at the problem of slow feature extraction speed of the KAZE algorithm and poor detection effect of moving targets

a moving target detection algorithm based on improved KAZE algorithm is proposed. On the basis of original KAZE algorithm

revise the iterative step size for solving the linear non-scale space

after solving the feature points displacement between adjacent frames in the feature matching stage for eliminating background. The experimental results show that the improved algorithm accelerates the solution of nonlinear scale space

extraction speed of the feature points is increased by 9. 3 %

and feature matching rate is improved by 1. 6 %. In moving targets detection of unmanned aerial vehicle（ UVA）

good effects are acquired.