Research On Morphological Feature Extraction Algorithm Of Optical Barefoot Image

Footprints have always occupied an important position in the criminal investigation field because of their extensiveness,uniqueness,stability and collectability.Among them,morphological features are the most important and frequently used features in footprint recognition.Footprint experts can calculate height,weight,gender and other identity information based on the morphological features of footprint images to narrow the scope of reconnaissance.In recent years,the field of footprint recognition has begun to be paid attention by more teams,and the research on the morphological features of footprints has become more and more in-depth.However,at this stage,the research on the morphological features of the footprint is based on complete and clear footprint images,and most of the footprint images extracted at the scene of the crime do not meet the quality requirements of the current stage of research.How to effectively extract morphological features from noisy or incomplete footprint images is an urgent problem to be solved.Aiming at the above problems,this paper adopts a two-stage morphological feature extraction algorithm.Firstly,the deep learning technology extracts the complete footprint contour image with high confidence from the noisy or incomplete image,and then automatically extracts the footprint morphological features from the contour image.The main research content of this paper can be divided into the following three aspects:(1)An optical barefoot image contour extraction algorithm based on GAN network is adopted,which can effectively extract the complete contour image from the low-quality barefoot image and eliminate the noise in the original image.The algorithm is implemented under the framework of Cycle GAN network,which has two generators and two discriminators in opposite directions.By converting the original image into a contour image and then into a barefoot image,it calculates the difference between the image obtained by the two conversions and the original image.So that the network has the ability to complete the incomplete image,and at the same time,it can erase the noise in the original image.At the same time,in order to improve the effect of network contour extraction,the spectral normalization mechanism and the self-attention mechanism are integrated in the network.The spectral normalization mechanism can limit the change range of the weight matrix during the training process and avoid the pattern collapse during the network training process.The self-attention mechanism can guide the network to pay attention to the corresponding information between footprint image structures and improve the network’s attention to key areas.Experiments show that the network can effectively solve the problem of contour extraction of low-quality original images.In order to quantify the ability of the network to extract contours,this paper also adopts a new quantitative evaluation method to evaluate the effect of the contour extraction algorithm of optical barefoot images.(2)A method to automatically extract footprint morphological features from contour images is presented,and a new set of morphological features for optical barefoot images is defined.The method determines a unified coordinate system by calculating the contour convex hull after extracting the edge contour,automatically marking the feature points in the unified coordinate system,and finally automatically extracting the footprint morphological features.It effectively avoids the influence of subjective factors on the extraction results when manually extracting morphological features.Through the identity retrieval experiment after screening the morphological features,the effectiveness of using the two-stage morphological feature extraction algorithm in this paper to extract morphological features is proved.(3)At present,in the field of application,the morphological features of footprint images generally need to be manually extracted,which is easily affected by the subjective emotions of the extractors.In order to eliminate the influence of this uncertain factor,this paper designed the optical barefoot image morphological feature extraction system,the system standardizes the scientific pre-processing of the extracted optical barefoot image,extracts the contour or extracts the morphological features,which can effectively reduce the error caused by human operation.