In order to solve the problems of current HU invariant moment matching

that the calculation is too large to match quickly

and the precision of large rotation and scale change matching is low

a fast image matching algorithm based on improved HU invariant moments is proposed. The algorithm uses a pyramid model

which proportionally reduces the resolution of the template image and the image to be matched

and propose an improved matching search strategy

which optimizes the starting match position and search path

reduces the computational complexity of template traversal. The algorithm proposes a method for constructing multi-scale templates

and use the similarity measure and connected domain algorithm to achieve the target accurate positioning and marking. The experimental results show that this algorithm has certain invariance properties in grayscale variation

translation

large rotation and scaling

the matching speed is improved greatly for large resolution image

and a good matching effect is obtained.