Study Of Pork With Skin Traceability Based On Pore-scale Features

Pork is one of the main meat varieties in China.Although the government has paid a lot of manpower and material resources to supervise all aspects of pork production,there have been a series of serious potk safety accidents in recent years.At present,the traceability of pork in our country mainly depends on the seal on pigskin and the information recorded by the staff of each transaction link,instead of depending on the characteristics of pork itself.As the characteristics of the pork with skin,the pores are ubiquitous in the pigskin,and can maintain the same character in the circulation process.This study is dedicated to the exploration and extraction of the unique feature information of the pores in the pigskin images and then based on these features to realize the matching between the local pigskin image and the whole pigskin image,and then to realize the traceability of the pork with skin.The main contents and conclusions of this paper are as follows.(1)Study on the extraction of pore features of high definition pigskin image.According to the target application scenario of this study,we went to the market to buy pigskin samples and then collected the high definition images of pigskin.According to the characteristics of the pores and the target application scenario,this study proposes a multi-direction pore feature extraction algorithm named MPSIFT(Multi-orientation Pore Scale Invariant Feature Transform)based on the PSIFT(Pore Scale Invariant Feature Transform)algorithm,which makes the pore feature of pigskin image have a certain rotation invariance.Then this study uses MPSIFT algorithm to detect and extract the pore features and generates at least one feature point descriptor vector for each pore feature point.(2)Study on the matching between local pigskin image and whole pigskin image.Based on the feature points extracted by MPSIFT algorithm from pigskin image,this paper studys the matching between local pigskin image and whole pigskin image.In the study,we use the vectorial angle as the similarity measure of the feature points in the pigskin image,and use the ratio of the angle between different feature points to measure the matching degree between the feature points.After matching the feature points of two pigskin images,this study filters the matched feature points by outlier detection so as to ensure the quality of matched points.(3)Based on the study on the matching between local pigskin image and whole pigskin image,this paper proposes a traceability technology of the pork with skin based on pore-scale features and then expounded the traceability principle and application scenarios.(4)Experiment and analysis of the feature extraction and matching of pigskin images.This study finds out the effect of different data characteristics on the matching of the pigskin images by experiments and finally achieves 94.14% matching accuracy between the local pigskin image and the whole pigskin image.The tracing success rate of the local pigskin image 86.73% is achieved when the tracing success results are reliable.