Research Of Anti-counterfeiting Algorithm Of Eriocheir Sinensis Based On Local Features Of Images

In recent years,with the rapid development of economy and society and the rise of e-commerce,and with the rise of logistics and other industries,the consumer market and consumer groups of Eriocheir sinensis are also in a growing trend.The traditional methods of river crab tracing and anti-counterfeiting usually use barcodes or QR codes that are manually tied to crab claws.Due to the mobility of the markers,the crab body is easily replaced or counterfeited.The germplasm traits such as bulges,trenches,ridges and teeth on the back shell of the crab change slightly due to the heterogeneity of the living environment,making the above-mentioned traits of the individual's carapace unique.Based on the uniqueness of the characteristics of the crab husk,this paper proposes an anti-counterfeiting algorithm based on the local features of the image,which solves the defects of the traditional crab anti-counterfeiting method and realize the purpose of food safety verification.The anti-counterfeiting algorithm of Eriocheir sinensis based on the local features of the image is studied around the uniqueness of the features of the back image of the crab.Firstly,the image segmentation technology is researched and analyzed.According to the prior knowledge that the back image is the largest connected domain in the whole image,a combination algorithm based on threshold,edge and region is adopted to realize the back images segmentation.Secondly,the classical image local feature detection algorithm is studied in detail,and the SURF(speeded up robust features)algorithm with better performance is selected to complete the feature detection of the back image.Thirdly,the feature point matching of the two images is completed by the two-way FLANN(fast library for approximate nearest neighbors)algorithm,which effectively reduces the "one-to-many" situation in the matching and reduces the number of mismatches.Fourthly,it is determined whether the matching pairs are mismatched by judging whether the position of matching pairs in two images is consistent with the nearest neighbor Euclidean distance and the next nearest neighbor Euclidean distance matching point.And the number of correct matching points is counted.Finally,the ratio of the correct matching number to the total matching number is taken as the similarity of the two images.If the similarity is greater than a certain threshold,the two images belong to the same crab,otherwise they are not considered to be the same crab.Experiments were carried out using real crab images,and the similarity threshold in the algorithm is determined by experimental data to be 0.7.The experimental results show that under similar image acquisition conditions,the similarity of the two back shell images belonging to the same crab is higher than the threshold value of 0.7,while the similarity of different crab back shell images is lower.Therefore,a database can be established for the collected the genuine crabs' images captured at the production site,and the algorithm of this paper can be used to match the crabs' images on the market with the images of database,thereby realizing the monomer identification and traceability of crabs,and achieving the purpose of anti-counterfeiting and food safety verification.