Scenic area locker management system and method based on face recognition

Abstract

The invention discloses a scenic spot locker management system and method based on face recognition, and relates to the technical field of lockers, and the management method comprises the following steps: setting a face recognition rule to carry out similarity evaluation on each recognition record, and generating a recognition log of each user; performing exception division on each identification record in any identification log; extracting the difference characteristics of the identification logs, and analyzing the occurrence frequency and the similarity evaluation condition of the difference characteristics in each identification log to obtain the influence degree of each difference characteristic; when the user carries out face recognition, a recognition record is generated, similarity evaluation is carried out, if the obtained similarity is lower than a similarity threshold value, related information of the user about different features is captured, a plurality of different features met by the user are obtained, the similarity is adjusted, and the user is subjected to face recognition; and if the adjusted similarity is higher than a set similarity threshold, providing an adjustment suggestion for the user and performing identification again.

Description

Technical Field

[0001] The present invention relates to the technical field of lockers, and in particular to a scenic area locker management system and method based on face recognition. Background Art

[0002] Facial recognition lockers are an intelligent storage solution that uses advanced facial recognition technology to achieve safe storage and access of items. It not only improves the convenience of the storage process, but also enhances the security of the stored items, and is gradually being used in various occasions.

[0003] Although compared with traditional recognition methods, it has the advantages of fast access and no reliance on identified objects, it also creates new problems. For example, when recognizing faces, it will be interfered by various factors including but not limited to light exposure and obstruction of facial decorations, resulting in face recognition failure. Once recognition fails, it may continue to fail, affecting the user's access efficiency and mood, and greatly reducing the user's trust in lockers. Summary of the invention

[0004] The purpose of the present invention is to provide a scenic spot locker management system and method based on face recognition to solve the problems raised in the prior art.

[0005] To achieve the above object, the present invention provides the following technical solution: a scenic spot locker management method based on face recognition, the management method comprising the following steps:

[0006] Step S100: Establish a face recognition management system in the lockers of the scenic area, record each face recognition process of the lockers of the scenic area, and obtain a number of recognition records; based on the facial feature comparison during the face recognition process, calculate the similarity of each recognition record, and determine the similarity threshold for judging whether the face recognition is successful; the recognition record stores information such as the timestamp, the facial feature set entered by the user, the currently captured user facial image, the recognition result, and the matching degree;

[0007] Step S200: Generate a recognition log for each user based on the facial features of each user stored in the face recognition management system; perform abnormal classification on each recognition record in any recognition log; extract the difference features of the recognition log by comparing the differences between the recognition records; and include all recognition records with the same set of facial features entered by the user in the recognition log;

[0008] Step S300: Analyze the occurrence frequency of the difference features in each recognition log and classify each difference feature; compare the difference features and similarity evaluation between each recognition record in each recognition log to obtain the influence degree of each difference feature on face recognition;

[0009] Step S400: Whenever a user performs face recognition, an identification record is generated and a similarity evaluation is performed. If the obtained similarity is lower than the similarity threshold, the user's relevant information on each difference feature is captured to obtain a number of difference features that the user satisfies. The similarity is adjusted according to the degree of influence of the several difference features. If the adjusted similarity is higher than the similarity threshold, adjustment suggestions are provided to the user and the recognition is performed again.

[0010] Furthermore, step S100 includes the following steps:

[0011] Step S101: extracting features from the user's facial image recorded when the user uses the locker in the scenic spot for the first time, obtaining the user's facial feature set and storing it in the face recognition management system; arbitrarily selecting a recognition record, comparing the user's facial image captured in the recognition record with the pre-stored facial feature set, and obtaining the similarity S of the recognition record;

[0012] Step S102: Obtain the recognition results recorded in each recognition record, and mark the recognition records that failed to be recognized with abnormal features; divide all recognition records according to whether there are abnormal feature marks to obtain a normal record set and an abnormal record set;

[0013] Step S103: Obtain the similarity of each identified record in the normal record set and the abnormal record set respectively, and extract the normal similarity threshold (S 1 ) min and the anomaly similarity threshold (S 2 ) max ; If (S 1 ) min ≤(S 2 ) max , then the normal similarity threshold (S 1 ) min As the similarity threshold for judging whether face recognition is successful, if (S 1 ) min >(S 2 ) max , then the abnormal similarity threshold (S 2 ) max As the similarity threshold for judging whether face recognition is successful; the similarity of the normal record set is the similarity range of successful recognition, and the similarity of the abnormal record set is the similarity range of failed recognition. By taking appropriate values ​​between the two similarity ranges, a reasonable similarity threshold can be obtained to judge whether face recognition is successful.

[0014] Further, step S200 includes the following steps:

[0015] Step S201: Obtain a facial feature set in each recognition record for feature comparison with the user's facial image. If there are two recognition records with the same facial feature set, aggregate the two recognition records to generate a recognition log; compare the facial feature sets of all recognition records to generate a recognition log corresponding to each user;

[0016] Step S202: arbitrarily selecting a recognition log, dividing each recognition record in the recognition log into a normal record set and an abnormal record set; extracting one recognition record from each of the two sets, extracting key features of the user's facial images captured in the two recognition records, and obtaining key feature sets of the two recognition records respectively;

[0017] Step S203: preset several appearance types for the facial feature sets stored in each recognition record, respectively compare the key features in the two key feature sets with various appearance types, and determine the appearance type corresponding to each key feature; summarize the several appearance types corresponding to the two key feature sets to obtain an appearance type set; the appearance features include the shape of the facial features, whether or not the user is wearing any decorations, etc. For example, if a user wears glasses when recording a facial image, wearing glasses can be used as an appearance feature of the user;

[0018] Step S204: arbitrarily select a type of appearance type from the appearance type set, select key features corresponding to the appearance type from the two key feature sets respectively, perform difference comparison on the two selected key features to obtain the difference degree between the two key features, set a difference degree threshold, if the difference degree exceeds the difference degree threshold, then set the appearance type corresponding to the two key features as a difference feature of the recognition log; perform difference comparison on each recognition record in the recognition log to generate a difference feature set of the recognition log.

[0019] Further, step S300 includes the following steps:

[0020] Step S301: randomly select a recognition log and extract a difference feature set of the recognition log, randomly select a difference feature from the difference feature set, and count the number of occurrences of the difference feature in the i-th recognition log as N. i , set the number of recognition records in the i-th recognition log to M i ; According to the formula:

[0021]

[0022] Where i and j are positive integers, i∈(1,a), j∈(1,a), a is the number of recognition logs stored in the face recognition management system, M jis the number of identification records in the jth identification log; the comprehensive occurrence frequency f of the difference feature is calculated ’ The comprehensive occurrence frequency reflects the occurrence frequency of a difference feature in each identification record, which is used for subsequent comparison with the occurrence frequency in an identification record to determine whether the occurrence frequency in an identification log is normal, which is helpful for determining the cause of the difference in the difference feature;

[0023] Step S302: Count the number of times the difference feature satisfies N in each recognition log. i The number of logs with the difference feature ≠ 0 is b, and the occurrence ratio of the difference feature in all the identified logs is calculated to be η = a / b; a threshold of occurrence ratio η is preset. max , if η≥η max , then the difference feature is marked as the first feature; if η<η max , then the abnormal frequency judgment value of the difference feature in each recognition log is obtained as f p =f ’ / η; obtain the occurrence frequency of the difference feature in the i-th recognition log as f i =N i / M i , when f is satisfied i ≥f p When f is satisfied, the difference feature is marked as a second feature. i <f p When the difference feature is , the third feature mark is added to the difference feature; the first feature mark indicates that the objective factors of the scenic spot lockers cause the recognition failure, such as light exposure, system fluctuations, etc.; the second feature mark indicates that the subjective factors of the user corresponding to the recognition log cause the recognition failure, such as changes in the user's facial decorations and appearance; the third feature mark indicates that the recognition failure occurs under special circumstances;

[0024] Step S303: arbitrarily select the kth difference feature from the difference feature set of the recognition log, and set the influence degree of the kth difference feature to x. k ; Select any one identification record from the identification log, obtain the feature mark corresponding to each difference feature in the identification record, according to the formula:

[0025]

[0026] Where k1, k2 and k3 are positive integers, and k1∈(1,c), k2∈(1,c), k3∈(1,c); Flag k1 is the feature flag corresponding to the k1th difference feature, IF() is the judgment function, IF(Flag k1=1) is used to determine whether the k1th difference feature has the first feature flag. If the first feature flag exists, then IF(Flag k1 =1)=1, otherwise, IF(Flag k1 =1)=0; Flag k2 is the feature flag corresponding to the k2th difference feature, IF(Flag k2 =2) is used to determine whether the k2th difference feature has a second feature marker. If the second feature marker exists, then IF(Flag k2 =2)=1, otherwise, IF(Flag k2 =2)=0; Flag k3 is the feature flag corresponding to the k3th difference feature, IF(Flag k3 =3) is used to determine whether the k3th difference feature has the third feature marker. If the third feature marker exists, then IF(Flag k3 =3)=1, otherwise, IF(Flag k3 =3)=0; x k1 is the influence degree of the k1th difference feature, x k2 is the influence degree of the k2th difference feature, x k3 is the influence degree of the k3th difference feature; f k2 is the occurrence frequency of the k2th difference feature in the recognition log, η k3 is the occurrence ratio of the k3th difference feature in all recognition logs; calculate the comprehensive influence degree Y of all difference features with the first feature mark in the recognition log 1 , the comprehensive influence degree Y of all differential features with the second feature marker 2 and the comprehensive influence degree Y of all differential features with the third characteristic marker 3 ;

[0027] Since the difference features are divided into three categories, the impact degree analysis methods of each difference feature are also different. Since the objective factors have problems in each recognition log, the impact degree can be directly obtained. The user's subjective factors are independent difference features in each recognition log, so it is necessary to analyze the impact degree in combination with the frequency of occurrence in the recognition log. Since the probability of occurrence of special difference features is small, it can be analyzed by comprehensive frequency of occurrence.

[0028] Step S304: randomly select one identification record from the normal record set and the abnormal record set of the identification log, and obtain the similarity of the two selected identification records respectively, according to the formula:

[0029]

[0030] Wherein, S1 is the similarity of the selected identification records in the normal record set, and S2 is the similarity of the selected identification records in the abnormal record set;

[0031] Step S305: Obtain the similarity difference between any two recognition records in the two record sets in the recognition log to determine the influence degree x k1 , x k2 and x k3 wherein the influence degree of each difference feature having the first characteristic mark and the second characteristic mark is calculated on an average value to obtain the average influence degree of each difference feature.

[0032] Further, step S400 includes the following steps:

[0033] Step S401: Perform face recognition on the current user, match it with the user information stored in the face recognition management system, generate a real-time recognition record, determine the recognition log where the real-time recognition record is located, and obtain the similarity of the real-time recognition record as S. now ;

[0034] Step S402: Set the similarity threshold to S ’ , if S now <S ’ , based on the preset attributes of different categories, the key information generated in the user face recognition process is extracted to obtain the key information corresponding to each attribute; the key information corresponding to each attribute in each recognition record of the recognition log is obtained, and the similarity of the two sets of key information is compared according to the attributes to obtain the difference feature set satisfied by the current user;

[0035] Step S403: Obtain the influence of each difference feature in the difference feature set in the recognition log, and set the influence of the dth difference feature as x. d , according to the formula:

[0036]

[0037] Wherein, w is the number of difference features of the difference feature set satisfied by the current user; the similarity S after the real-time recognition record is adjusted is calculated. new If S new >S ’, then provide adjustment suggestions to the user and perform re-recognition; when the first recognition fails, the user's facial features can be analyzed to determine whether there are certain features that affect facial recognition. If corresponding features exist, the impact of the corresponding features can be checked and the user's facial recognition results can be adjusted. If the adjusted similarity meets the conditions, it means that the reason for the user's recognition failure is caused by the influence of some factors, which helps the user to actively check the relevant reasons and re-recognize to ensure recognition efficiency.

[0038] In order to better implement the above method, a scenic spot locker management system based on face recognition is also proposed. The management system includes a face recognition evaluation module, a difference feature analysis module, a feature impact analysis module, and a face recognition real-time analysis module;

[0039] The face recognition evaluation module is used to establish a face recognition management system in the lockers of the scenic area, record each face recognition process of the lockers in the scenic area, and obtain a number of recognition records; based on the facial feature comparison during the face recognition process, calculate the similarity of each recognition record and determine the similarity threshold for judging whether the face recognition is successful;

[0040] The difference feature analysis module is used to generate a recognition log for each user according to the facial features of each user stored in the face recognition management system; to perform abnormal classification on each recognition record in any recognition log; and to extract the difference features of the recognition log by comparing the differences between the recognition records;

[0041] The feature impact analysis module is used to analyze the occurrence frequency of the difference features in each recognition log and classify each difference feature; compare the difference features and similarity evaluation between each recognition record in each recognition log to obtain the influence of each difference feature on face recognition;

[0042] The real-time face recognition analysis module is used to generate a recognition record and perform a similarity assessment every time a user performs face recognition. If the similarity obtained is lower than the similarity threshold, the user's relevant information on each difference feature is captured to obtain a number of difference features that the user satisfies; the similarity is adjusted according to the degree of influence of the several difference features. If the adjusted similarity is higher than the similarity threshold, adjustment suggestions are provided to the user and re-recognition is performed.

[0043] Further, the face recognition evaluation module includes a recognition record construction unit and a similarity evaluation unit;

[0044] The recognition record construction unit is used to establish a face recognition management system in the lockers of the scenic area, record each face recognition process of the lockers of the scenic area, and obtain a number of recognition records; the similarity evaluation unit is used to calculate the similarity of each recognition record based on the facial feature comparison during the face recognition process, and determine the similarity threshold for judging whether the face recognition is successful.

[0045] Further, the difference feature analysis module includes an abnormal record division unit and a difference feature extraction unit;

[0046] The abnormal record classification unit is used to generate a recognition log for each user according to the facial features of each user stored in the face recognition management system; to perform abnormal classification on each recognition record in any recognition log; and the difference feature extraction unit is used to extract the difference features of the recognition log by comparing the differences between the recognition records.

[0047] Further, the feature impact analysis module includes a difference feature division unit and an impact degree analysis unit;

[0048] The difference feature division unit is used to analyze the occurrence frequency of the difference features in each recognition log and divide each difference feature; the impact degree analysis unit is used to compare the difference features and similarity evaluation between each recognition record in each recognition log to obtain the impact degree of each difference feature on face recognition.

[0049] Further, the face recognition real-time analysis module includes a feature capturing unit and an adjustment suggestion providing unit;

[0050] The feature capture unit is used to generate a recognition record and perform a similarity evaluation every time a user performs face recognition. If the similarity obtained is lower than the similarity threshold, the user's relevant information on each difference feature is captured to obtain a number of difference features that the user satisfies; the adjustment suggestion providing unit is used to adjust the similarity according to the influence of the several difference features. If the adjusted similarity is higher than the similarity threshold, the user is provided with adjustment suggestions and re-recognition is performed.

[0051] Compared with the prior art, the present invention has the following beneficial effects:

[0052] 1. The present invention helps to identify various factors that lead to facial recognition failure by analyzing the differences between historical recognition records, and can help users make timely adjustments to avoid frequent recognition failures, thereby improving users' access efficiency and access mood, and enhancing users' trust in lockers;

[0053] 2. The present invention analyzes the influence of each difference feature, takes into account the difference between objective difference features and subjective difference features, and performs respective analyses to obtain a more accurate influence of each difference feature, thereby helping to effectively analyze the similarity of user face recognition in the subsequent process, obtain a more accurate similarity, and improve the recognition efficiency of the user;

[0054] 3. When the user fails to identify for the first time, the present invention can actively help the user analyze the cause of the identification failure and predict the similarity after solving the cause of the failure. If the conditions for successful identification are met, the user is reminded to make adjustments, thereby helping the user to improve the accuracy of successful identification in the next identification and improve the user's access efficiency. BRIEF DESCRIPTION OF THE DRAWINGS

[0055] Figure 1 A schematic diagram of the steps of a scenic spot locker management method based on face recognition;

[0056] Figure 2 The figure is a structural diagram of a scenic spot locker management system based on face recognition. DETAILED DESCRIPTION

[0057] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0058] Example: Figure 1 to Figure 2 As shown, the present invention provides a scenic spot locker management method based on face recognition, and the management method comprises the following steps:

[0059] Step S100: Establish a face recognition management system in the lockers of the scenic area, record each face recognition process of the lockers of the scenic area, and obtain a number of recognition records; based on the facial feature comparison during the face recognition process, calculate the similarity of each recognition record, and determine the similarity threshold for judging whether the face recognition is successful;

[0060] Wherein, step S100 includes the following steps:

[0061] Step S101: extracting features from the user's facial image recorded when the user uses the locker in the scenic spot for the first time, obtaining the user's facial feature set and storing it in the face recognition management system; arbitrarily selecting a recognition record, comparing the user's facial image captured in the recognition record with the pre-stored facial feature set, and obtaining the similarity S of the recognition record;

[0062] Step S102: Obtain the recognition results recorded in each recognition record, and mark the recognition records that failed to be recognized with abnormal features; divide all recognition records according to whether there are abnormal feature marks to obtain a normal record set and an abnormal record set;

[0063] Step S103: Obtain the similarity of each identified record in the normal record set and the abnormal record set respectively, and extract the normal similarity threshold (S 1 ) min and the anomaly similarity threshold (S 2 ) max ; If (S 1 ) min ≤(S 2 ) max , then the normal similarity threshold (S 1 ) min As the similarity threshold for judging whether face recognition is successful, if (S 1 ) min >(S 2 ) max , then the abnormal similarity threshold (S 2 ) max As a similarity threshold to determine whether face recognition is successful;

[0064] Example 1: Set the similarity range obtained in the normal record set to (90%, 100%). If the similarity range of the abnormal record set is (0, 80%), the similarity threshold for judging whether face recognition is successful is 80%. If the similarity range of the abnormal record set is (0, 92%), the similarity threshold for judging whether face recognition is successful is 90%.

[0065] Step S200: generating a recognition log for each user according to the facial features of each user stored in the face recognition management system; performing abnormal classification on each recognition record in any recognition log; extracting difference features of the recognition log by comparing the differences between the recognition records;

[0066] Wherein, step S200 includes the following steps:

[0067] Step S201: Obtain a facial feature set in each recognition record for feature comparison with the user's facial image. If there are two recognition records with the same facial feature set, aggregate the two recognition records to generate a recognition log; compare the facial feature sets of all recognition records to generate a recognition log corresponding to each user;

[0068] Step S202: arbitrarily selecting a recognition log, dividing each recognition record in the recognition log into a normal record set and an abnormal record set; extracting one recognition record from each of the two sets, extracting key features of the user's facial images captured in the two recognition records, and obtaining key feature sets of the two recognition records respectively;

[0069] Step S203: Preset a number of appearance types for the facial feature sets stored in each recognition record, respectively compare the key features in the two key feature sets with the various appearance types, and determine the appearance type corresponding to each key feature; summarize the several appearance types corresponding to the two key feature sets to obtain an appearance type set;

[0070] Step S204: arbitrarily select a type of appearance type from the appearance type set, select key features corresponding to the appearance type from the two key feature sets respectively, perform difference comparison on the two selected key features to obtain the difference degree between the two key features, set a difference degree threshold, if the difference degree exceeds the difference degree threshold, then set the appearance type corresponding to the two key features as a difference feature of the recognition log; perform difference comparison on each recognition record in the recognition log to generate a difference feature set of the recognition log.

[0071] Step S300: Analyze the occurrence frequency of the difference features in each recognition log and classify each difference feature; compare the difference features and similarity evaluation between each recognition record in each recognition log to obtain the influence degree of each difference feature on face recognition;

[0072] Wherein, step S300 includes the following steps:

[0073] Step S301: randomly select a recognition log and extract a difference feature set of the recognition log, randomly select a difference feature from the difference feature set, and count the number of occurrences of the difference feature in the i-th recognition log as N. i , set the number of recognition records in the i-th recognition log to M i ; According to the formula:

[0074]

[0075] Where i and j are positive integers, i∈(1,a), j∈(1,a), a is the number of recognition logs stored in the face recognition management system, M j is the number of identification records in the jth identification log; the comprehensive occurrence frequency f of the difference feature is calculated ’ ;

[0076] Example 2: Assume that the number of occurrences of a difference feature in three recognition logs is 5 times, 6 times and 3 times respectively, and the number of recognition records in the three recognition logs is 10 times, 15 times and 10 times respectively, and the comprehensive occurrence frequency of the difference feature is f ’ =14 / 35=40%;

[0077] Step S302: Count the number of times the difference feature satisfies N in each recognition log. i The number of logs with the difference feature ≠ 0 is b, and the occurrence ratio of the difference feature in all the identified logs is calculated to be η = a / b; a threshold of occurrence ratio η is preset. max , if η≥η max , then the difference feature is marked as the first feature; if η<η max , then the abnormal frequency judgment value of the difference feature in each recognition log is obtained as f p =f ’ / η; obtain the occurrence frequency of the difference feature in the i-th recognition log as f i =N i / M i , when f is satisfied i ≥f p When f is satisfied, the difference feature is marked as a second feature. i <f p When the difference feature is marked with a third feature;

[0078] Example 3: The number of times a difference feature exists in each identification log is set to 9 times, and the number of identification logs is 10 times. The occurrence ratio of the difference feature is obtained to be 90%, and the occurrence ratio threshold is set to 70%. Therefore, the difference feature is marked as the first feature, which is the identification failure caused by objective factors;

[0079] Step S303: arbitrarily select the kth difference feature from the difference feature set of the recognition log, and set the influence degree of the kth difference feature to x. k ; Select any one identification record from the identification log, obtain the feature mark corresponding to each difference feature in the identification record, according to the formula:

[0080]

[0081] Where k1, k2 and k3 are positive integers, and k1∈(1,c), k2∈(1,c), k3∈(1,c); Flag k1 is the feature flag corresponding to the k1th difference feature, IF() is the judgment function, IF(Flag k1=1) is used to determine whether the k1th difference feature has the first feature marker. If the first feature marker exists, then IF(Flag k1 =1)=1, otherwise, IF(Flag k1 =1)=0; Flag k2 is the feature flag corresponding to the k2th difference feature, IF(Flag k2 =2) is used to determine whether the k2th difference feature has a second feature marker. If the second feature marker exists, then IF(Flag k2 =2)=1, otherwise, IF(Flag k2 =2)=0; Flag k3 is the feature flag corresponding to the k3th difference feature, IF(Flag k3 =3) is used to determine whether the k3th difference feature has the third feature marker. If the third feature marker exists, then IF(Flag k3 =3) = 1, otherwise, IF(Flag k3 =3)=0; x k1 is the influence degree of the k1th difference feature, x k2 is the influence degree of the k2th difference feature, x k3 is the influence degree of the k3th difference feature; f k2 is the occurrence frequency of the k2th difference feature in the recognition log, η k3 is the occurrence ratio of the k3th difference feature in all recognition logs; calculate the comprehensive influence degree Y of all difference features with the first feature mark in the recognition log 1 , the comprehensive influence degree Y of all differential features with the second feature marker 2 and the comprehensive influence degree Y of all differential features with the third characteristic marker 3 ;

[0082] Step S304: randomly select one identification record from the normal record set and the abnormal record set of the identification log, and obtain the similarity of the two selected identification records respectively, according to the formula:

[0083]

[0084] Wherein, S1 is the similarity of the selected identification records in the normal record set, and S2 is the similarity of the selected identification records in the abnormal record set;

[0085] Step S305: Obtain the similarity difference between any two recognition records in the two record sets in the recognition log to determine the influence degree x k1 , x k2 and x k3wherein the influence degree of each difference feature having the first characteristic mark and the second characteristic mark is calculated on an average value to obtain the average influence degree of each difference feature.

[0086] Step S400: Whenever a user performs face recognition, a recognition record is generated and a similarity evaluation is performed. If the obtained similarity is lower than the similarity threshold, relevant information of the user on each difference feature is captured to obtain a number of difference features satisfied by the user; the similarity is adjusted according to the influence of the several difference features. If the adjusted similarity is higher than the similarity threshold, an adjustment suggestion is provided to the user and recognition is performed again;

[0087] Wherein, step S400 includes the following steps:

[0088] Step S401: Perform face recognition on the current user, match it with the user information stored in the face recognition management system, generate a real-time recognition record, determine the recognition log where the real-time recognition record is located, and obtain the similarity of the real-time recognition record as S. now ;

[0089] Step S402: Set the similarity threshold to S ’ , if S now <S ’ , based on the preset attributes of different categories, the key information generated in the user face recognition process is extracted to obtain the key information corresponding to each attribute; the key information corresponding to each attribute in each recognition record of the recognition log is obtained, and the similarity of the two sets of key information is compared according to the attributes to obtain the difference feature set satisfied by the current user;

[0090] Step S403: Obtain the influence of each difference feature in the difference feature set in the recognition log, and set the influence of the dth difference feature as x. d , according to the formula:

[0091]

[0092] Wherein, w is the number of difference features of the difference feature set satisfied by the current user; the similarity S after the real-time recognition record is adjusted is calculated. new If S new >S ’ , then provide adjustment suggestions to the user and perform re-identification;

[0093] Example 4: Set the similarity obtained by the user's first face recognition to 70%, and the similarity threshold to 90%. Therefore, it is necessary to extract the difference features of the user to obtain the user wearing a mask and wearing glasses, where the influence of wearing a mask is 20%, and the influence of wearing glasses is 10%. The adjusted similarity is calculated to be 80%×(1+20%)×(1+10%)=92.4%. Because 92.4%>90%, the system reminds the user to take off the mask and glasses and perform the test again.

[0094] A scenic spot locker management system based on face recognition, the management system includes a face recognition evaluation module, a difference feature analysis module, a feature impact analysis module, and a face recognition real-time analysis module;

[0095] The face recognition evaluation module is used to establish a face recognition management system in the lockers of the scenic area, record each face recognition process of the lockers in the scenic area, and obtain a number of recognition records; based on the facial feature comparison during the face recognition process, calculate the similarity of each recognition record and determine the similarity threshold for judging whether the face recognition is successful;

[0096] The difference feature analysis module is used to generate a recognition log for each user according to the facial features of each user stored in the face recognition management system; to perform abnormal classification on each recognition record in any recognition log; and to extract the difference features of the recognition log by comparing the differences between the recognition records;

[0097] The feature impact analysis module is used to analyze the occurrence frequency of the difference features in each recognition log and classify each difference feature; compare the difference features and similarity evaluation between each recognition record in each recognition log to obtain the influence of each difference feature on face recognition;

[0098] The real-time face recognition analysis module is used to generate a recognition record and perform a similarity assessment every time a user performs face recognition. If the similarity obtained is lower than the similarity threshold, the user's relevant information on each difference feature is captured to obtain a number of difference features that the user satisfies; the similarity is adjusted according to the degree of influence of the several difference features. If the adjusted similarity is higher than the similarity threshold, adjustment suggestions are provided to the user and re-recognition is performed.

[0099] Among them, the face recognition evaluation module includes a recognition record construction unit and a similarity evaluation unit;

[0100] The recognition record construction unit is used to establish a face recognition management system in the lockers of the scenic area, record each face recognition process of the lockers of the scenic area, and obtain a number of recognition records; the similarity evaluation unit is used to calculate the similarity of each recognition record based on the facial feature comparison during the face recognition process, and determine the similarity threshold for judging whether the face recognition is successful.

[0101] Among them, the difference feature analysis module includes an abnormal record division unit and a difference feature extraction unit;

[0102] The abnormal record classification unit is used to generate a recognition log for each user according to the facial features of each user stored in the face recognition management system; to perform abnormal classification on each recognition record in any recognition log; and the difference feature extraction unit is used to extract the difference features of the recognition log by comparing the differences between the recognition records.

[0103] Among them, the feature impact analysis module includes a difference feature division unit and an impact degree analysis unit;

[0104] The difference feature division unit is used to analyze the occurrence frequency of the difference features in each recognition log and divide each difference feature; the impact degree analysis unit is used to compare the difference features and similarity evaluation between each recognition record in each recognition log to obtain the impact degree of each difference feature on face recognition.

[0105] Among them, the face recognition real-time analysis module includes a feature capture unit and an adjustment suggestion providing unit;

[0106] The feature capture unit is used to generate a recognition record and perform a similarity evaluation every time a user performs face recognition. If the similarity obtained is lower than the similarity threshold, the user's relevant information on each difference feature is captured to obtain a number of difference features that the user satisfies; the adjustment suggestion providing unit is used to adjust the similarity according to the influence of the several difference features. If the adjusted similarity is higher than the similarity threshold, the user is provided with adjustment suggestions and re-recognition is performed.

[0107] It will be apparent to those skilled in the art that the invention is not limited to the details of the exemplary embodiments described above and that the invention can be implemented in other specific forms without departing from the spirit or essential features of the invention. Therefore, the embodiments should be considered exemplary and non-limiting in all respects, and the scope of the invention is defined by the appended claims rather than the foregoing description, and it is intended that all variations falling within the meaning and scope of the equivalent elements of the claims be included in the invention. Any reference numeral in a claim should not be considered as limiting the claim to which it relates.

Claims

1. A scenic area locker management method based on face recognition, characterized in that: The management method comprises the following steps: Step S100: Establish a face recognition management system in the lockers of the scenic area, record each face recognition process of the lockers of the scenic area, and obtain a number of recognition records; based on the facial feature comparison during the face recognition process, calculate the similarity of each recognition record, and determine the similarity threshold for judging whether the face recognition is successful; Step S200: generating a recognition log for each user according to the facial features of each user stored in the face recognition management system; performing abnormal classification on each recognition record in any recognition log; extracting difference features of the recognition log by comparing the differences between the recognition records; Step S300: Analyze the occurrence frequency of the difference features in each recognition log and classify each difference feature; compare the difference features and similarity evaluation between each recognition record in each recognition log to obtain the influence degree of each difference feature on face recognition; Step S400: Whenever a user performs face recognition, an identification record is generated and a similarity evaluation is performed. If the obtained similarity is lower than the similarity threshold, the user's relevant information on each difference feature is captured to obtain a number of difference features that the user satisfies. The similarity is adjusted according to the degree of influence of the several difference features. If the adjusted similarity is higher than the similarity threshold, adjustment suggestions are provided to the user and the recognition is performed again.

2. A scenic area locker management method based on face recognition according to claim 1, characterized in that: The step S100 includes the following steps: Step S101: extracting features from the user's facial image recorded when the user uses the locker in the scenic spot for the first time, obtaining the user's facial feature set and storing it in the face recognition management system; arbitrarily selecting a recognition record, comparing the user's facial image captured in the recognition record with the pre-stored facial feature set, and obtaining the similarity S of the recognition record; Step S102: Obtain the recognition results recorded in each recognition record, and mark the recognition records that failed to be recognized with abnormal features; divide all recognition records according to whether there are abnormal feature marks to obtain a normal record set and an abnormal record set; Step S103: Obtain the similarity of each identified record in the normal record set and the abnormal record set respectively, and extract the normal similarity threshold (S1) with the smallest value in the normal record set. min and the anomaly similarity threshold (S2) with the largest value in the anomaly record set max ; if (S1) min ≤(S2) max , then the normal similarity threshold (S1) min As the similarity threshold for judging whether face recognition is successful, if (S1) min >(S2) max , then the abnormal similarity threshold (S2) max As the similarity threshold for judging whether face recognition is successful.

3. A scenic area locker management method based on face recognition according to claim 2, characterized in that: The step S200 includes the following steps: Step S201: Obtain a facial feature set in each recognition record for feature comparison with the user's facial image. If there are two recognition records with the same facial feature set, aggregate the two recognition records to generate a recognition log; compare the facial feature sets of all recognition records to generate a recognition log corresponding to each user; Step S202: arbitrarily selecting a recognition log, dividing each recognition record in the recognition log into a normal record set and an abnormal record set; extracting one recognition record from each of the two sets, extracting key features of the user's facial images captured in the two recognition records, and obtaining key feature sets of the two recognition records respectively; Step S203: Preset a number of appearance types for the facial feature sets stored in each recognition record, respectively compare the key features in the two key feature sets with the various appearance types, and determine the appearance type corresponding to each key feature; summarize the several appearance types corresponding to the two key feature sets to obtain an appearance type set; Step S204: arbitrarily select a type of appearance type from the appearance type set, select key features corresponding to the appearance type from the two key feature sets respectively, perform difference comparison on the two selected key features to obtain the difference degree between the two key features, set a difference degree threshold, if the difference degree exceeds the difference degree threshold, then set the appearance type corresponding to the two key features as a difference feature of the recognition log; perform difference comparison on each recognition record in the recognition log to generate a difference feature set of the recognition log.

4. A scenic area locker management method based on face recognition according to claim 3, characterized in that: The step S300 includes the following steps: Step S301: randomly select a recognition log and extract a difference feature set of the recognition log, randomly select a difference feature from the difference feature set, and count the number of occurrences of the difference feature in the i-th recognition log as N. i , set the number of recognition records in the i-th recognition log to M i ; According to the formula: Where i and j are positive integers, i∈(1,a), j∈(1,a), a is the number of recognition logs stored in the face recognition management system, M j is the number of identification records in the jth identification log; the comprehensive occurrence frequency f of the difference feature is calculated ’ ; Step S302: Count the number of times the difference feature satisfies N in each recognition log. i The number of logs with the difference feature ≠ 0 is b, and the occurrence ratio of the difference feature in all the identified logs is calculated to be η = a / b; a threshold of occurrence ratio η is preset. max , if η≥η max , then the difference feature is marked as the first feature; if η<η max , then the abnormal frequency judgment value of the difference feature in each recognition log is obtained as f p =f ’ / η; obtain the occurrence frequency of the difference feature in the i-th recognition log as f i =N i / M i , when f is satisfied i ≥f p When f is satisfied, the difference feature is marked as a second feature. i <f p When the difference feature is marked with a third feature; Step S303: arbitrarily select the kth difference feature from the difference feature set of the recognition log, and set the influence degree of the kth difference feature to x. k ; Select any one identification record from the identification log, obtain the feature mark corresponding to each difference feature in the identification record, according to the formula: Where k1, k2 and k3 are positive integers, and k1∈(1,c), k2∈(1,c), k3∈(1,c); Flag k1 is the feature flag corresponding to the k1th difference feature, IF() is the judgment function, IF(Flag k1 =1) is used to determine whether the k1th difference feature has the first feature marker. If the first feature marker exists, then IF(Flag k1 =1)=1, otherwise, IF(Flag k1 =1)=0; Flag k2 is the feature flag corresponding to the k2th difference feature, IF(Flag k2 =2) is used to determine whether the k2th difference feature has a second feature marker. If the second feature marker exists, then IF(Flag k2 =2)=1, otherwise, IF(Flag k2 =2)=0; Flag k3 is the feature flag corresponding to the k3th difference feature, IF(Flag k3 =3) is used to determine whether the k3th difference feature has the third feature marker. If the third feature marker exists, then IF(Flag k3 =3) = 1, otherwise, IF(Flag k3 =3)=0; x k1 is the influence degree of the k1th difference feature, x k2 is the influence degree of the k2th difference feature, x k3 is the influence degree of the k3th difference feature; f k2 is the occurrence frequency of the k2th difference feature in the recognition log, η k3 is the occurrence ratio of the k3th difference feature in all recognition logs; calculate and obtain the comprehensive influence degree Y1 of all difference features with the first feature mark, the comprehensive influence degree Y2 of all difference features with the second feature mark, and the comprehensive influence degree Y3 of all difference features with the third feature mark in the recognition log; Step S304: randomly select one identification record from the normal record set and the abnormal record set of the identification log, and obtain the similarity of the two selected identification records respectively, according to the formula: Wherein, S1 is the similarity of the selected identification records in the normal record set, and S2 is the similarity of the selected identification records in the abnormal record set; Step S305: Obtain the similarity difference between any two recognition records in the two record sets in the recognition log to determine the influence degree x k1 , x k2 and x k3 wherein the influence degree of each difference feature having the first characteristic mark and the second characteristic mark is calculated on an average value to obtain the average influence degree of each difference feature.

5. A scenic area locker management method based on face recognition according to claim 4, characterized in that: The step S400 includes the following steps: Step S401: Perform face recognition on the current user, match it with the user information stored in the face recognition management system, generate a real-time recognition record, determine the recognition log where the real-time recognition record is located, and obtain the similarity of the real-time recognition record as S. now ; Step S402: Set the similarity threshold to S ’ , if S now <S ’ , based on the preset attributes of different categories, the key information generated in the user face recognition process is extracted to obtain the key information corresponding to each attribute; the key information corresponding to each attribute in each recognition record of the recognition log is obtained, and the similarity of the two sets of key information is compared according to the attributes to obtain the difference feature set satisfied by the current user; Step S403: Obtain the influence of each difference feature in the difference feature set in the recognition log, and set the influence of the dth difference feature as x. d , according to the formula: Wherein, w is the number of difference features of the difference feature set satisfied by the current user; the similarity S after the real-time recognition record is adjusted is calculated. new If S new >S ’ , then provide adjustment suggestions to the user and perform re-identification.

6. A scenic spot locker management system, used to implement a scenic spot locker management method based on face recognition as described in any one of claims 1 to 5, characterized in that: The management system includes a face recognition evaluation module, a difference feature analysis module, a feature impact analysis module, and a face recognition real-time analysis module; The face recognition evaluation module is used to establish a face recognition management system in the lockers of the scenic area, record each face recognition process of the lockers of the scenic area, and obtain a number of recognition records; based on the facial feature comparison during the face recognition process, calculate the similarity of each recognition record, and determine the similarity threshold for judging whether the face recognition is successful; The difference feature analysis module is used to generate a recognition log for each user according to the facial features of each user stored in the face recognition management system; to perform abnormal classification on each recognition record in any recognition log; and to extract the difference features of the recognition log by comparing the differences between the recognition records; The feature impact analysis module is used to analyze the occurrence frequency of the difference features in each recognition log and classify each difference feature; Compare the difference features and similarity evaluations between the recognition records in each recognition log to obtain the influence of each difference feature on face recognition; The face recognition real-time analysis module is used to generate a recognition record and perform a similarity evaluation every time a user performs face recognition. If the obtained similarity is lower than the similarity threshold, the user's relevant information on each difference feature is captured to obtain a number of difference features satisfied by the user; the similarity is adjusted according to the influence degree of the several difference features. If the adjusted similarity is higher than the similarity threshold, adjustment suggestions are provided to the user and re-recognition is performed.

7. A scenic area locker management system according to claim 6, characterized in that: The face recognition evaluation module includes a recognition record construction unit and a similarity evaluation unit; The recognition record construction unit is used to establish a face recognition management system in the lockers of the scenic area, record each face recognition process of the lockers of the scenic area, and obtain a plurality of recognition records; the similarity evaluation unit is used to calculate the similarity of each recognition record based on the facial feature comparison during the face recognition process, and determine the similarity threshold for judging whether the face recognition is successful.

8. A scenic area locker management system according to claim 6, characterized in that: The difference feature analysis module includes an abnormal record division unit and a difference feature extraction unit; The abnormal record classification unit is used to generate a recognition log for each user according to the facial features of each user stored in the face recognition management system; to perform abnormal classification on each recognition record in any recognition log; and the difference feature extraction unit is used to extract the difference features of the recognition log by comparing the differences between the recognition records.

9. A scenic area locker management system according to claim 6, characterized in that: The feature impact analysis module includes a difference feature division unit and an impact degree analysis unit; The difference feature division unit is used to analyze the occurrence frequency of the difference features in each recognition log and divide each difference feature; the impact degree analysis unit is used to compare the difference features and similarity evaluation between each recognition record in each recognition log to obtain the impact degree of each difference feature on face recognition.

10. A scenic area locker management system according to claim 6, characterized in that: The face recognition real-time analysis module includes a feature capture unit and an adjustment suggestion providing unit; The feature capture unit is used to generate an identification record and perform a similarity evaluation each time the user performs face recognition. If the similarity obtained is lower than the similarity threshold, the user's relevant information about each difference feature is captured to obtain a number of difference features that the user satisfies. The adjustment suggestion providing unit is used to adjust the similarity according to the influence of the several difference features. If the adjusted similarity is higher than the similarity threshold, the user is provided with adjustment suggestions and re-identification is performed.

Images