Multi-scale pattern recognition method and system

Abstract

The invention relates to the technical field of pattern recognition, and discloses a multi-scale pattern recognition method, which comprises the following steps: acquiring a to-be-recognized pattern, converting the to-be-recognized pattern into a pattern grey-scale map, and carrying out binarization processing on the pattern grey-scale map by utilizing a local maximum between-class variance method to obtain a binarized pattern; performing detail enhancement processing on the binarized graph by using an adaptive graph enhancement algorithm; performing decomposition processing on the graph after detail enhancement by using a graph decomposition algorithm, and decomposing contour and texture scale information of the graph into different graphs; fusing the graphs of different scales by using a multi-scale fusion algorithm to obtain a fused graph; and taking the fused graph as the input of a graph recognition model, and performing graph recognition on the fused graph by using the graph recognition model. The invention further provides a multi-scale pattern recognition system. According to the invention, multi-scale pattern recognition is realized.

Description

technical field [0001] The invention relates to the technical field of pattern recognition, in particular to a multi-scale pattern recognition method and system. Background technique [0002] Vision is an important means for people to perceive the external environment and plays a vital role in people's cognition of the world. Therefore, various images that can be intuitively felt by the human eye can be seen everywhere in people's life and work. With the development of science and technology, people's pursuit of high-definition visual experience is more urgent. [0003] Traditional image multi-scale segmentation techniques mainly include the following three types: the first is threshold-based segmentation, the second is region-based segmentation, and the third is graph theory-based segmentation. The influence of noise in the threshold method is large, which makes it difficult to accurately segment the image edges of multi-scale images, which affects the accuracy of segmenta...

AI Technical Summary

Problems solved by technology

The influence of noise in the threshold method is large, which makes it difficult to accurately segment the image edges of multi-scale images, which affects the accuracy of segmentation; while the region-based segmentation method will cause over-segmentation due to the noise on the image and the local discontinuity of the image. Problem; The segmentation method based on graph theory needs to select the number of segmentation blocks in advance, which has high complexity in time and space

Images