Fast curve matching for tattoo recognition and identification

Abstract

Some implementations of the systems and techniques described include a tattoo recognition system that is capable of improving the accuracy and efficiency associated with matching edge curves extracted from tattoo images, which are typically performed during a tattoo matching operation. The system can perform a matching operation in two stages—a feature extraction stage and a matching stage. In the first stage, the system extracts one or more edge curves from the tattoo image. In the second stage, the system performs matching using a two-step comparison that compares features of the extracted edge curves.

What is claimed is: 
     
         1 . A method comprising:
 obtaining data indicating (i) a first set of line segment sequences in a first set of polygonal representations generated for a search tattoo image and (ii) a second set of line segment sequences in a second set of polygonal representations generated for a reference tattoo image;   determining a set of best-matched line segment sequence pairs between the first set of line segment sequences and the second set of line segment sequences;   determining, from among the first set of polygonal representations and the second set of polygonal representations, a best-matched polygonal representation pair based on the set of best-matched line segment sequence pairs; and   computing a similarity score between the search tattoo image and the reference tattoo image based on the best-matched polygonal representation pair.   
     
     
         2 . The method of  claim 1 , wherein:
 the first set of polygonal representations are generated based on one or more search edge curves extracted from the search tattoo image; and   the second set of polygonal representations are generated based on one or more reference edge curves extracted from the reference tattoo image.   
     
     
         3 . The method of  claim 1 , wherein each best-matched line segment sequence pair included in the set of best-matched line segment sequence pairs comprises a particular line segment sequence that is detected in both the first set of polygonal representations and the second set of polygonal representations. 
     
     
         4 . The method of  claim 1 , wherein computing the similarity score comprises:
 determining a similarity between a search polygonal representation included in the best-matched polygonal representation pair and a reference polygonal representation included in the best-matched polygonal representation pair; and   computing a value of the similarity score based on the similarity determined between the search polygonal representation included in the best-matched polygonal representation pair and the reference polygonal representation included in the best-matched polygonal representation pair.   
     
     
         5 . The method of  claim 1 , wherein determining the best-matched polygonal representation pair comprises:
 computing, for each line segment sequence pair included in the set of best-matched line segment sequence pairs, a similarity score between a first line segment sequence from the first set of line segment sequences and a second line segment sequence from the second set of line segment sequences;   identifying, based on the similarity scores computed for the set of best-matched line segment sequence pairs, a particular line segment sequence pair with a confidence score having a highest relative value; and   identifying a particular polygonal representation pair that includes the particular line segment sequence pair.   
     
     
         6 . The method of  claim 1 , further comprising:
 determining a first set of feature vectors for the first set of polygonal representations;   determining a second set of feature vectors for the second set of polygonal representations; and   wherein the set of best-matched line segment sequence pairs is determined based at least on a comparison of the first set of feature vectors and the second set of feature vectors.   
     
     
         7 . The method of  claim 1 , wherein the similarity score indicates a likelihood that the search tattoo image and the reference tattoo image each include a same tattoo. 
     
     
         8 . A system comprising:
 one or more computing devices; and   one or more storage devices storing instructions that, when executed by the one or more computing devices, cause the one or more computing devices to perform operations comprising:   obtaining data indicating (i) a first set of line segment sequences in a first set of polygonal representations generated for a search tattoo image and (ii) a second set of line segment sequences in a second set of polygonal representations generated for a reference tattoo image;   determining a set of best-matched line segment sequence pairs between the first set of line segment sequences and the second set of line segment sequences;   determining, from among the first set of polygonal representations and the second set of polygonal representations, a best-matched polygonal representation pair based on the set of best-matched line segment sequence pairs; and   computing a similarity score between the search tattoo image and the reference tattoo image based on the best-matched polygonal representation pair.   
     
     
         9 . The system of  claim 8 , wherein:
 the first set of polygonal representations are generated based on one or more search edge curves extracted from the search tattoo image; and   the second set of polygonal representations are generated based on one or more reference edge curves extracted from the reference tattoo image.   
     
     
         10 . The system of  claim 8 , wherein each best-matched line segment sequence pair included in the set of best-matched line segment sequence pairs comprises a particular line segment sequence that is detected in both the first set of polygonal representations and the second set of polygonal representations. 
     
     
         11 . The system of  claim 8 , wherein computing the similarity score comprises:
 determining a similarity between a search polygonal representation included in the best-matched polygonal representation pair and a reference polygonal representation included in the best-matched polygonal representation pair; and   computing a value of the similarity score based on the similarity determined between the search polygonal representation included in the best-matched polygonal representation pair and the reference polygonal representation included in the best-matched polygonal representation pair.   
     
     
         12 . The system of  claim 8 , wherein determining the best-matched polygonal representation pair comprises:
 computing, for each line segment sequence pair included in the set of best-matched line segment sequence pairs, a similarity score between a first line segment sequence from the first set of line segment sequences and a second line segment sequence from the second set of line segment sequences;   identifying, based on the similarity scores computed for the set of best-matched line segment sequence pairs, a particular line segment sequence pair with a confidence score having a highest relative value; and   identifying a particular polygonal representation pair that includes the particular line segment sequence pair.   
     
     
         13 . The system of  claim 8 , wherein the operations further comprise:
 determining a first set of feature vectors for the first set of polygonal representations;   determining a second set of feature vectors for the second set of polygonal representations; and   wherein the set of best-matched line segment sequence pairs is determined based at least on a comparison of the first set of feature vectors and the second set of feature vectors.   
     
     
         14 . The system of  claim 8 , wherein the similarity score indicates a likelihood that the search tattoo image and the reference tattoo image each include a same tattoo. 
     
     
         15 . At least one non-transitory computer-readable storage device storing instructions that, when executed by one or more processors, cause the one or more processors to perform operations comprising:
 obtaining data indicating (i) a first set of line segment sequences in a first set of polygonal representations generated for a search tattoo image and (ii) a second set of line segment sequences in a second set of polygonal representations generated for a reference tattoo image;   determining a set of best-matched line segment sequence pairs between the first set of line segment sequences and the second set of line segment sequences;   determining, from among the first set of polygonal representations and the second set of polygonal representations, a best-matched polygonal representation pair based on the set of best-matched line segment sequence pairs; and   computing a similarity score between the search tattoo image and the reference tattoo image based on the best-matched polygonal representation pair.   
     
     
         16 . The non-transitory computer-readable storage device of  claim 15 , wherein:
 the first set of polygonal representations are generated based on one or more search edge curves extracted from the search tattoo image; and   the second set of polygonal representations are generated based on one or more reference edge curves extracted from the reference tattoo image.   
     
     
         17 . The non-transitory computer-readable storage device of  claim 15 , wherein each best-matched line segment sequence pair included in the set of best-matched line segment sequence pairs comprises a particular line segment sequence that is detected in both the first set of polygonal representations and the second set of polygonal representations. 
     
     
         18 . The non-transitory computer-readable storage device of  claim 15 , wherein computing the similarity score comprises:
 determining a similarity between a search polygonal representation included in the best-matched polygonal representation pair and a reference polygonal representation included in the best-matched polygonal representation pair; and   computing a value of the similarity score based on the similarity determined between the search polygonal representation included in the best-matched polygonal representation pair and the reference polygonal representation included in the best-matched polygonal representation pair.   
     
     
         19 . The non-transitory computer-readable storage device of  claim 15 , wherein determining the best-matched polygonal representation pair comprises:
 computing, for each line segment sequence pair included in the set of best-matched line segment sequence pairs, a similarity score between a first line segment sequence from the first set of line segment sequences and a second line segment sequence from the second set of line segment sequences;   identifying, based on the similarity scores computed for the set of best-matched line segment sequence pairs, a particular line segment sequence pair with a confidence score having a highest relative value; and   identifying a particular polygonal representation pair that includes the particular line segment sequence pair.   
     
     
         20 . The non-transitory computer-readable storage device of  claim 15 , wherein the operations further comprise:
 determining a first set of feature vectors for the first set of polygonal representations;   determining a second set of feature vectors for the second set of polygonal representations; and   wherein the set of best-matched line segment sequence pairs is determined based at least on a comparison of the first set of feature vectors and the second set of feature vectors.