Affine invariant feature extraction methods have been one of the key issues of computer vision research.This paper proposes a normalized histogram algorithm based on the density function of multi-scale autoconvolution transform and researches the normalization method of the density function

and constructs a method to extract the histogram affine invariant feature

and implements affine invariant pattern recognition based on multi-scale autoconvolution normalized histograms.Simulation results show that this algorithm has good adaptability for a certain range of noise

partial occlusion

illumination and view angle variance.The recognition rate of this algorithm is superior to multi-scale autoconvolution and other multi-scale autoconvolution based histogram algorithms

especially under a variety of environments.