# US 11,443,552 B2 · Image pattern similarity calculation device and recognition device > Agents: this data is available as JSON from the patentgrep API and MCP server — see https://patentgrep.com/llms.txt - Number: US 11,443,552 B2 - Document ID: US11443552B2 - Status: Granted - Kind: B2 - Application number: 15/258,230 - Filed: 2016-09-07 - Published: 2017-09-21 - Granted: 2022-09-13 - Earliest claimed: 2016-03-17 - Assignees: Kabushiki Kaisha Toshiba - Applicants: Kabushiki Kaisha Toshiba - Inventors: Tomoyuki Shibata - Web page: https://patentgrep.com/patent/US11443552B2 ## CPC - G06V 40/172: {Classification, e.g. identification} - G06F 18/22: Matching criteria, e.g. proximity measures - G06V 10/761: {Proximity, similarity or dissimilarity measures} - G06V 30/242: Division of the character sequences into groups prior to recognition; Selection of dictionaries - G06V 40/168: {Feature extraction; Face representation} - G06V: IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING - G06F: ELECTRIC DIGITAL DATA PROCESSING (computer systems based on specific computational models G06N) ## Abstract According to an embodiment, a calculation device includes a registerer, a first calculator, a receiver, and a second calculator. The registerer registers a plurality of patterns. The first calculator calculates a degree of closeness between the registered patterns. The receiver receives an input pattern. The second calculator calculates a first similarity value between the input pattern and a first registered pattern among the plurality of registered patterns, calculates second similarity values between the input pattern and one or more second registered patterns having a neighbor relationship with the first registered pattern among the plurality of registered patterns, and calculates a combined similarity value in which the first similarity value and the one or more second similarity values are combined. ## Claims ### Claim 1 A calculation device comprising: a processor configured to: register a plurality of registered patterns comprising a set of a feature value extracted from an image, each pattern corresponding to an object target of a plurality of targets, the object target being extracted from a moving image; calculate a degree of closeness between the registered patterns using a normalized cross-correlation; receive an input pattern; set a registered pattern which is not yet set as a first registered pattern among the registered patterns, as the first registered pattern; generate a category including the first registered pattern among the registered patterns and one or more second registered patterns having a highest degree of closeness with the first registered pattern among the registered patterns; calculate a first similarity value between the input pattern and the first registered pattern in the category; calculate one or more second similarity values between the input pattern and the one or more second registered patterns in the category, the one or more second similarity values being different from the first similarity value; calculate a combined similarity value in which the first similarity value and the one or more second similarity values are combined as a similarity value of the category; perform the setting of the registered pattern, the generating of the category, the calculating of the first similarity value, the calculating of the one or more second similarity values, and the calculating of the combined similarity value repeatedly until any registered pattern not set as the first registered pattern does not remain in the plurality of registered patterns; recognize that the input pattern belongs to the category based on the plurality of combined similarity values; and output the combined similarity values calculated. ### Claim 2 (depends on 1) The device according to claim 1, wherein the processor is further configured to combine the first similarity value, the one or more second similarity values, and degrees of closeness between the one or more second registered patterns and the first registered pattern to calculate a second combined similarity value as the combined similarity value. ### Claim 3 (depends on 1) The device according to claim 1, wherein the processor is configured to further calculate degrees of closeness between the registered patterns and the input pattern using a normalized cross-correlation, calculate one or more third similarity values between the first registered pattern and one or more third registered patterns having a highest degree of closeness with the input pattern among the registered patterns; combine the first similarity value, the one or more second similarity values, and the one or more third similarity values to calculate a third combined similarity value as the combined similarity value; and perform the calculating of the one or more third similarity values and the combining repeatedly until any registered pattern not set as the first registered pattern does not remain in the plurality of registered patterns. ### Claim 4 (depends on 3) The device according to claim 3, wherein the processor is further configured to combine the first similarity value, the one or more second similarity values, the one or more third similarity values, the degrees of closeness between the one or more second registered patterns and the first registered pattern, and the degrees of closeness between the one or more third registered patterns and the input pattern to calculate a fourth combined similarity value as the combined similarity value. ### Claim 5 (depends on 1) The device according to claim 1, wherein the processor is further configured to set a plurality of reference patterns, calculate degrees of closeness between the registered patterns and the plurality of reference patterns using a normalized cross-correlation, and calculate, as the one or more second similarity values, similarity values between the input pattern and one or more second registered patterns and first reference patterns having a highest degree of closeness with the first registered pattern among the registered patterns and the plurality of reference patterns. ### Claim 6 (depends on 5) The device according to claim 5, wherein the processor is further configured to combine the first similarity value, the one or more second similarity values, and the degrees of closeness between the first registered pattern and the one or more second registered patterns and first reference patterns to calculate a fifth combined similarity value as the combined similarity value. ### Claim 7 (depends on 5) The device according to claim 5, wherein the processor is configured to further calculate degrees of closeness between the registered patterns and the input pattern and degrees of closeness between the plurality or reference patterns and the input pattern using a normalized cross-correlation, calculate one or more third similarity values between the first registered pattern and one or more third registered patterns and second reference patterns having a highest degree of closeness with the input pattern among the registered patterns and the plurality of reference patterns, and combine the first similarity value, the one or more second similarity values, and the one or more third similarity values to calculate a six combined similarity value as the combined similarity value. ### Claim 8 (depends on 7) The device according to claim 7, wherein the processor is further configured to combine the first similarity value, the one or more second similarity values, the one or more third similarity values, the degrees of closeness between the first registered pattern and the one or more second registered patterns and first reference patterns, and the degrees of closeness between the input pattern and the one or more third registered patterns and second reference patterns to calculate a seventh combined similarity value as the combined similarity value. ### Claim 9 A recognition device comprising: a processor configured to: register a plurality of patterns comprising a set of a feature value extracted from an image, each pattern corresponding to an object target of a plurality of targets, the object target being extracted from a moving image; calculate a degree of closeness between the registered patterns using a normalized cross-correlation; receive an input pattern; set a registered pattern which is not yet set as a first registered pattern among the registered patterns, as the first registered pattern; generate a category including a first registered pattern among the registered patterns and one or more second registered patterns having a highest degree of closeness with the first registered pattern among the registered patterns; calculate, for each of the registered patterns, a first similarity value between the input pattern and the first registered pattern in the category; calculate one or more second similarity values between the input pattern and the one or more second registered patterns in the category, the second similarity values being different from the first similarity value; calculate a combined similarity value in which the first similarity value and the one or more second similarity values are combined as a similarity value of the category; perform the setting of the first registered pattern, the generating of the category, the calculating of the first similarity value, the calculating of the one or more second similarity values, and the calculating of the combined similarity value repeatedly until any registered pattern not set as the first registered pattern does not remain in the plurality of registered patterns; recognize that the input pattern belongs to the category based on the plurality of combined similarity values; and output a category to which the input pattern belongs. ### Claim 10 (depends on 9) The device according to claim 9, wherein the processor is further configured to combine the first similarity value, the one or more second similarity values, and degrees of closeness between the one or more second registered patterns and the first registered pattern to calculate a second combined similarity value as the combined similarity value. ### Claim 11 (depends on 9) The device according to claim 9, wherein the processor is configured to further calculate degrees of closeness between the registered patterns and the input pattern using a normalized cross-correlation, calculate one or more third similarity values between the first registered pattern and one or more third registered patterns having a highest degree of closeness with the input pattern among the registered patterns; combine the first similarity value, the one or more second similarity values, and the one or more third similarity values to calculate a third combined similarity value as the combined similarity value; and perform the calculating of the one or more third similarity values and the combining repeatedly until any registered pattern not set as the first registered pattern does not remain in the plurality of registered patterns. ### Claim 12 (depends on 11) The device according to claim 11, wherein the processor is further configured to combine the first similarity value, the one or more second similarity values, the one or more third similarity values, the degrees of closeness between the one or more second registered patterns and the first registered pattern, and the degrees of closeness between the one or more third registered patterns and the input pattern to calculate a fourth combined similarity value as the combined similarity value. ### Claim 13 (depends on 9) The device according to claim 9, wherein the processor is further configured to set a plurality of reference patterns, calculate degrees of closeness between the registered patterns and the plurality of reference patterns using a normalized cross-correlation, and calculate, as the one or more second similarity values, similarity values between the input pattern and one or more second registered patterns and first reference patterns having a highest degree of closeness with the first registered pattern among the registered patterns and the plurality of reference patterns. ### Claim 14 (depends on 13) The device according to claim 13, wherein the processor is further configured to combine the first similarity value, the one or more second similarity values, and the degrees of closeness between the first registered pattern and the one or more second registered patterns and first reference patterns to calculate a fifth combined similarity value as the combined similarity value. ### Claim 15 (depends on 13) The device according to claim 13, wherein the processor is configured to further calculate degrees of closeness between the registered patterns and the input pattern and degrees of closeness between the plurality of reference patterns and the input pattern using a normalized cross-correlation, calculate one or more third similarity values between the first registered pattern and one or more third registered patterns and second reference patterns having a highest degree of closeness with the input pattern among the registered patterns and the plurality of reference patterns, and combine the first similarity value, the one or more second similarity values, and the one or more third similarity values to calculate a sixth combined similarity value as the combined similarity value. ### Claim 16 (depends on 15) The device according to claim 15, wherein the processor is further configured to combine the first similarity value, the one or more second similarity values, the one or more third similarity values, the degrees of closeness between the first registered pattern and the one or more second registered patterns and first reference patterns, and the degrees of closeness between the input pattern and the one or more third registered patterns and second reference patterns to calculate a seventh combined similarity value as the combined similarity value. ### Claim 17 (depends on 9) The device according to claim 9, wherein the processor is further configured to recognize that the input pattern belongs to a category of the first registered pattern in which the combined similarity value having the highest similarity value is calculated among the plurality of combined similarity values. ### Claim 18 (depends on 9) The device according to claim 9, wherein the processor is further configured to recognize that the input pattern belongs to a category of one or more first registered patterns in which the combined similarity value having a similarity value equal to or greater than a threshold value is calculated among the plurality of combined similarity values.