http://rdf.ncbi.nlm.nih.gov/pubchem/patent/CN-107330416-B
Outgoing Links
Predicate | Object |
---|---|
classificationCPCInventive | http://rdf.ncbi.nlm.nih.gov/pubchem/patentcpc/G06F18-285 http://rdf.ncbi.nlm.nih.gov/pubchem/patentcpc/G06F18-214 http://rdf.ncbi.nlm.nih.gov/pubchem/patentcpc/G06V20-53 |
classificationIPCInventive | http://rdf.ncbi.nlm.nih.gov/pubchem/patentipc/G06K9-62 http://rdf.ncbi.nlm.nih.gov/pubchem/patentipc/G06K9-00 |
filingDate | 2017-07-10-04:00^^<http://www.w3.org/2001/XMLSchema#date> |
grantDate | 2019-08-09-04:00^^<http://www.w3.org/2001/XMLSchema#date> |
publicationDate | 2019-08-09-04:00^^<http://www.w3.org/2001/XMLSchema#date> |
publicationNumber | CN-107330416-B |
titleOfInvention | A Pedestrian Re-Identification Method Based on Discriminative Structure Measure Learning |
abstract | The embodiment of the present invention discloses a pedestrian re-identification method based on discriminant structure measure learning, which includes: dividing each training image into K sub-regions and extracting features; Region metric; learn non-corresponding sub-regions of similar pedestrian image pairs to obtain weak out-of-class sub-region metric; learn corresponding and non-corresponding sub-regions of non-similar pedestrian image pairs to obtain out-of-class sub-region metric; get mapping matrix based on three metrics H; calculate the Mahalanobis distance of the corresponding sub-regions of two pedestrian images; train to obtain the weight of the sub-region; obtain the Mahalanobis distance between the test image and the corresponding sub-region of each training image, and then obtain the test image and all training images. Based on the similarity score, the pedestrian re-identification result of the test image is obtained. The invention fully excavates the structural information of the image and automatically finds the discriminative sub-region, thereby improving the correct rate of pedestrian re-identification and matching. |
priorityDate | 2017-07-10-04:00^^<http://www.w3.org/2001/XMLSchema#date> |
type | http://data.epo.org/linked-data/def/patent/Publication |
Incoming Links
Predicate | Subject |
---|---|
isDiscussedBy | http://rdf.ncbi.nlm.nih.gov/pubchem/compound/CID25572 http://rdf.ncbi.nlm.nih.gov/pubchem/substance/SID415713197 |
Total number of triples: 15.