A progressive refinement method for shoe print pattern extraction combined with prior knowledge

Abstract

The invention discloses a progressive refinement method for extracting shoe print patterns combined with prior knowledge. The steps include: S1 collecting original on-site shoe print image information, and using superpixel segmentation to obtain image sub-regions with consistent features; S2 pairing the shoe prints Roughly mark the pattern image; S3 judge the pattern area and the background area on the shoe print pattern image, and obtain the initial shoe print pattern extraction result P 1 ; S4 for the initial shoe print pattern extraction result P 1 Perform iterative refinement and correction to obtain the accurate shoe print pattern extraction result P; S5 performs illumination correction and pattern enhancement processing on the accurate shoe print pattern extraction result P; S6 uses the OTSU algorithm to binarize the processed image pattern extraction result , to get the final binary image of the shoe print pattern. The invention calculates the regional similarity based on the priori knowledge of the color and direction characteristics of the shoe print pattern, judges the pattern and the background, improves the accuracy of the pattern extraction, and further promotes the application and development of the footprint recognition technology.

Description

technical field [0001] The invention relates to the technical field of image processing, in particular to a progressive refinement extraction method for shoe print patterns combined with prior knowledge. Background technique [0002] At present, there are mainly two ways to extract shoe print patterns: [0003] (1) Manual extraction, mainly by experienced professionals using image processing software to segment the pattern area of ​​the shoe print image and extract the shoe print pattern. [0004] (2) Automatic extraction method: Liu Jiankang uses the fuzzy C-means clustering algorithm (FCM) to segment the shoe print image, and proposes the maximum inter-class variance method based on the two-dimensional attribute histogram to extract the shoe print pattern. Yang Fang classified the on-site shoe print images by background, and used different filtering methods to denoise different backgrounds. The method of multi-scale top-hat transformation is used to enhance the contrast ...

AI Technical Summary

Problems solved by technology

[0005] The above methods can achieve on-site extraction of shoe print patterns to a certain extent, but there are corresponding problems as follows: (1) The manual extraction method has a large workload and low efficiency, and there is semantic ambiguity and subjective (2) The automatic extraction method does not use the prior knowledge of the shoe print pattern and the user's semantic information, and the extracted pattern may be missing or contain a lot of background interference.

These problems will seriously affect the efficiency of public security personnel in solving cases, and will also restrict the application and development of footprint recognition technology.

Images