Multi-modal biological feature fusion security check leak detection prevention system

Abstract

The invention, which relates to the technical field of artificial intelligence and civil aviation informatization, discloses a multi-modal biological feature fusion security check leak detection prevention system comprising a risk sensing unit, a data acquisition unit, a real-time monitoring unit, a data analysis unit and a verification unit. According to the invention, active identification and marking of high-risk targets are realized in the isolation area through intelligent behavior analysis, accurate shunting and enhanced verification are carried out on passengers entering the isolation area, and key targets are ensured to be controlled; and finally, carrying out continuous identity verification and dynamic tracking at key nodes such as a security check channel and a gallery bridge by means of a multi-mode biological feature fusion technology. The closed-loop process not only realizes the conversion from passive prevention to active intervention, but also greatly improves the real-time discovery capability and accurate interception efficiency of leak detection behaviors through a layer-by-layer progressive prevention and control mechanism, effectively avoids the problem that non-security-checked passengers enter an isolation area and even boarding, and improves the safety of the passengers. And the reliability and the comprehensive operation efficiency of the security check system are comprehensively enhanced.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of artificial intelligence and civil aviation informatization, in particular to a multi-modal biological feature fusion security check anti-missed detection system. BACKGROUND

[0002] Security check is the first physical defense line in public safety management and is also a key link for preventing major security incidents. Once dangerous goods, fake identity personnel and suspicious behavior targets are missed, it is extremely likely to cause irreparable consequences for subsequent links. Therefore, preventing missed detection is directly related to the effectiveness and integrity of the entire security system. In the existing civil aviation security check system, there is still a risk that individual passengers do not go through the complete security check process before entering the security check channel. Especially during peak passenger flow, due to abnormal behaviors such as passenger trailing and mixed walking, manual monitoring is difficult to discover in time, which can easily lead to missed detection. On the other hand, even if the passengers have completed the security check, they may still walk into the wrong corridor bridge or enter the non-corresponding boarding gate in the corridor bridge area. The existing system lacks dynamic collection, analysis and comparison of multi-modal features, making it difficult to identify and intervene in time, which poses a security risk.

[0003] In view of the above technical defects, a solution is proposed. SUMMARY

[0004] The purpose of the present application is to solve the technical problem that during peak passenger flow, due to abnormal behaviors such as passenger trailing and mixed walking, manual monitoring is difficult to discover in time, leading to the fact that the passing personnel do not go through the complete security check process before entering the security check channel.

[0005] In order to achieve the above purpose, the present application adopts the following technical scheme: a multi-modal biological feature fusion security check anti-missed detection system, comprising a data acquisition unit, a risk perception unit, a real-time monitoring unit, a data analysis unit and a verification unit. The data acquisition unit is used to acquire and verify the identity information and photo information of the passing personnel at the security check verification station. If the verification is passed, the state of the passing personnel is updated to have been checked, the security check state information is generated and sent to the risk perception unit. At the same time, the identity information and photo information of the checked personnel are integrated to generate a security check subarea face database and sent to the data analysis unit. Further, the identity information and photo information of the passing personnel are acquired and verified at the security check verification station, and the specific process is as follows: The identity information stored in the identity card of the passing personnel is read by the built-in identity card reading module in the specified sensing area of the people's card integrated machine. The identity information includes name, gender, nationality, date of birth, address, identity card number, issuing authority, validity period and standard certificate photo. The high-definition camera built in the human and certificate integrated machine detects the live body of the passing person, and the detection actions include nodding, shaking, opening mouth and blinking; the live body detection passes, and the real-time front face photo of the passing person is captured; The human and certificate integrated machine built-in face comparison algorithm extracts the feature points of the real-time front face photo and the standard certificate photo, compares the feature points, and calculates the similarity; wherein the feature points include the position, contour and distance of the facial features; A similarity threshold Xs is set, if the calculated similarity score is greater than or equal to the threshold Xs, it is determined that the verification is passed; otherwise, it is determined that the verification is not passed, and the re-verification is started; The similarity threshold Xs is determined based on historical sample data.

[0006] The risk perception unit receives security check state information, collects real-time video data of the passing person through high-precision wide-angle cameras in the security check channel and in the boarding bridge, updates and locates the corresponding passing person in the real-time video data according to the security check state information, analyzes the behavior of the passing person in the isolation area, calculates the missed detection coefficient of the passing person, and generates comprehensive data by integration and sends it to the verification unit, The isolation area is a restricted area from the security check to the boarding gate. Further, the missed detection coefficient is calculated to generate label information, including the following steps: S301. Based on the real-time video data, a multi-modal feature vector is extracted, including the following steps: Based on the OpenPose+CNN algorithm, a multi-modal feature extraction model is constructed to extract a multi-modal feature vector group, the input of the multi-modal feature extraction model is real-time video data, and the output is a behavior trajectory vector and an action feature vector; S302. Based on the behavior trajectory vector and the action feature vector, a behavior label is generated, including the standing time, the behavior switching frequency and the trailing mark; The standing time represents the time of the user staying in place in the security check channel in a non-queue state; The behavior switching frequency represents the number of times of switching behavior types of the user in a unit of time, wherein the behavior types include static, walking, looking at the mobile phone, checking the signboard and looking around; The trailing mark calculates the space trajectory of each person in the waiting area by using the laser camera, and through time-lag synchronous analysis, if the trajectory of the later passing person is synchronized with the trajectory of the former passing person, it is recorded as 1, otherwise it is recorded as 0; S303. Based on the behavior label, an abnormal behavior vector is constructed, and the missed detection coefficient is calculated; The abnormal behavior vector format is: U={standing time, behavior switching frequency, trailing mark}. The missed detection coefficient The calculation formula is: ; ; ; Wherein, is the length of stay, is the behavior switching frequency, M is the tail mark, is the cross feature factor; is the weight coefficient, which is obtained by training based on historical experimental data; The higher the value of the missed detection coefficient , the greater the risk of missed detection; Real-time monitoring unit, for collecting multi-modal biological feature information of passing personnel through intelligent devices deployed in security check channels and corridor areas, generating dynamic data sets through integration and sending to data analysis unit; Further, the intelligent device includes a long-focus face camera array, an inertial measurement unit, and an edge computing gateway; the dynamic data set includes face geometry, texture features, and iris texture features; the edge computing gateway performs real-time preprocessing on the dynamic data set; Data analysis unit, for receiving security check subarea face library and dynamic data set and processing, extracting multi-modal biological features of passing personnel from the dynamic data set and comparing with corresponding features in the security check subarea face library, generating comparison results and sending to verification unit; Further, generating the comparison results includes the following steps: Based on OpenCV algorithm, multi-modal biological feature extraction is performed on the dynamic data set, including face geometry, texture features and iris texture features, to generate comparison vector Bd=(Mb,Wl,Hm); At the same time, the same multi-modal biological feature extraction is performed on the security check subarea face library to generate comparison object vector Bx=(MbX,WlX,HmX); Wherein, Mb and MbX are face geometry similarity vectors, Wl and WlX are texture feature similarity vectors, and Hm and HmX are iris feature hamming distance vectors; Based on the comparison vector Bd=(Mb,Wl,Hm)and the comparison object vector Bx=(MbX,WlX,HmX), the difference coefficient Cy is calculated, and the specific steps are as follows: ; Wherein, is the amplification coefficient of the face geometry similarity vector, is the amplification coefficient of the texture feature similarity vector, An amplification coefficient of the iris feature Hamming distance vector; An amplification coefficient of the face geometry similarity vector An amplification coefficient of the texture feature similarity vector An amplification coefficient of the iris feature Hamming distance vector According to the discrimination ability score of each modal feature in the historical security comparison data, the amplification coefficient of each modal feature is determined according to the normalized proportion of the discrimination ability score of the modal feature and all modal features; if the collection quality of a modal feature is lower than a preset quality threshold, the amplification coefficient of the modal feature is automatically reduced, and the amplification coefficients of other modal features are proportionally increased; Difference thresholds cy1 and cy2 are set, and cy1 < cy2; If Cy > cy2, the comparison result is consistent with the person and the certificate; If cy1 < Cy ≤ cy2, the comparison result is abnormal in the comparison of the person and the certificate; If Cy ≤ cy1, the comparison result is inconsistent with the person and the certificate; The verification unit is used to receive the comprehensive data and the comparison result, and to confirm whether the passing person is a missed detection user according to the comprehensive data and the comparison result, and to perform secondary verification on the passing person when it is confirmed that there is a missed detection risk; Further, according to the comprehensive data and the comparison result, it is confirmed whether the passing person is a missed detection user, and when it is confirmed that there is a missed detection risk, secondary verification is performed on the passing person, including the following steps: S601. In the isolation area, according to the missed detection coefficient in the comprehensive data A missed detection coefficient threshold Lj is set, and when the missed detection coefficient of the passing person ≥ Lj, the passing person is determined to be a high missed detection risk object, and secondary verification and correct path guidance are performed on the passing person; When <Lj, the passing person is determined to be a normal user; S602. In the security channel, when the comparison result of the passing person is inconsistent with the person and the certificate, it is determined that the passing person has missed detection, and a re-security process is triggered; When the comparison result of the passing person is abnormal in the comparison of the person and the certificate, it is determined that the passing person has a missed detection risk, and secondary verification is performed on the passing person; if the secondary verification result is consistent with the person and the certificate, the passing person is determined to be a normal security user; otherwise, the passing person is determined to have missed detection and a re-security process is triggered; When the comparison result of the passing person is consistent with the person and the certificate, the passing person is directly determined to be a normal security user; In the gallery bridge channel, when the comparison result of the passing person is inconsistent with the person and the certificate, it is determined that the passing person has walked wrong gallery bridge, and correct path guidance is performed; If the comparison result of the pedestrian's ID card is abnormal, it is determined that the pedestrian is at risk of taking the wrong bridge and a second verification is performed on the pedestrian. If the second verification result is consistent with the ID card, the pedestrian is considered a normal user. Otherwise, the pedestrian is determined to have taken the wrong bridge and is guided to the correct path. When the comparison result of the passerby is consistent with the identity card, the passerby is directly identified as a normal user.

[0007] In summary, due to the adoption of the above technical solution, the beneficial effects of the present invention are: This multimodal biometric fusion security inspection and missed detection prevention system first uses intelligent behavioral analysis to actively identify and mark high-risk targets in the isolation area, and strengthens the verification of passengers entering the isolation area to ensure that major risk targets are under control; finally, with the help of multimodal biometric fusion technology, continuous identity verification and dynamic tracking are carried out at key nodes such as security inspection channels and jet bridges; this closed-loop process not only realizes the transformation from passive prevention to active intervention, but also greatly improves the real-time detection capability and precise interception efficiency of missed detection behaviors through a progressive prevention and control mechanism, while effectively avoiding flight delays caused by passengers going to the wrong boarding gate, and comprehensively enhancing the reliability, intelligence level and comprehensive operational efficiency of the security inspection system. BRIEF DESCRIPTION OF THE DRAWINGS

[0008] Figure 1 This is a structural diagram of the security inspection and missed detection prevention system based on multimodal biometric fusion. DETAILED DESCRIPTION

[0009] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. 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 making creative efforts are within the scope of protection of the present invention.

[0010] Example: Figure 1 As shown, the multimodal biometric fusion security inspection and missed detection prevention system includes a data collection unit, a risk perception unit, a real-time monitoring unit, a data analysis unit, and a verification unit; The data collection unit is used to collect the identity information and photo information of the passers-by at the security check verification counter and verify them. If the verification is passed, the passer-by's status is updated to "passed security check", and the security check status information is generated and sent to the risk perception unit. At the same time, the identity information and photo information of the passers-by are integrated to generate a security check zone face database and send it to the data analysis unit; At the security checkpoint, the identity information and photo of the person passing through are collected and verified. The specific process is as follows: The identity information stored in the identity card of the passing person is read by the identity card reading module built in the designated sensing area of the human and certificate integrated machine, and the identity information includes name, gender, nationality, birth date, address, identity card number, issuing authority, validity period, and standard certificate photo; The passing person is detected by the high-definition camera built in the human and certificate integrated machine, and the detection actions include nodding, shaking, opening the mouth, and blinking; the real-time front face photo of the passing person is captured by the live body detection; The feature points of the real-time front face photo and the standard certificate photo are extracted by the face comparison algorithm built in the human and certificate integrated machine, and the similarity is calculated; wherein, the feature points include the position, contour and distance of the facial features; A similarity threshold Xs is set, if the calculated similarity score is greater than or equal to the threshold Xs, it is determined that the verification is passed; otherwise, it is determined that the verification is not passed, and the re-verification is started; The similarity threshold Xs is determined based on historical sample data.

[0011] The risk perception unit is used to receive security check state information, collect real-time video data of the passing person through high-precision wide-angle cameras in the security check channel and the boarding bridge, and complete state update and positioning of the corresponding passing person in the real-time video data according to the security check state information, analyze the behavior of the passing person in the isolation area, calculate the missed detection coefficient of the passing person, and generate comprehensive data by integration and send it to the verification unit, The isolation area is a restricted area from the security check to the boarding gate; The missed detection coefficient is calculated to generate the annotation information, including the following steps: S301. Based on the real-time video data, a multi-modal feature vector is extracted, including the following steps: Based on the OpenPose+CNN algorithm, a multi-modal feature extraction model is constructed to extract a multi-modal feature vector group, and the input of the multi-modal feature extraction model is the real-time video data, and the output is the behavior trajectory vector and the action feature vector; S302. Based on the behavior trajectory vector and the action feature vector, a behavior label is generated, including the standing time, the behavior switching frequency, and the trailing mark; The standing time represents the time of the user staying in place in the security check channel in a non-queue state; The behavior switching frequency represents the number of times the user switches the behavior type in a unit of time, wherein the behavior type includes static, walking, looking at the mobile phone, checking the signboard, looking around, and turning back; The trailing mark calculates the spatial trajectory of each person in the waiting area by using the laser camera, and through time-lag synchronous analysis, if the trajectory of the later passing person is synchronized with the trajectory of the previous passing person, it is recorded as 1, otherwise it is recorded as 0. S303. Based on the behavior label, an abnormal behavior vector is constructed, and a missed detection coefficient is calculated; The abnormal behavior vector format is: U={standing time, behavior switching frequency, tailing mark}; Missed detection coefficient The calculation formula is: ; ; ; Among them, is the standing time, is the behavior switching frequency, and M is the tailing mark, is the cross feature factor; is the weight coefficient, which is obtained by training based on historical experimental data; The higher the value of the missed detection coefficient , the greater the risk of missed detection; A real-time monitoring unit is configured to collect multi-modal biological feature information of the passing person through intelligent devices deployed in the security check channel and the corridor bridge area, generate a dynamic data set by integration, and send it to the data analysis unit; The intelligent device includes a long-focus face camera array, an inertial measurement unit, and an edge computing gateway. The dynamic data set includes face geometry, texture features, and iris texture features. The edge computing gateway performs real-time preprocessing on the dynamic data set; The data analysis unit is configured to receive the security check partition face library and the dynamic data set and process them, extract the multi-modal biological features of the passing person from the dynamic data set, compare them with the corresponding features in the security check partition face library, generate a comparison result, and send it to the verification unit; Generating a comparison result includes the following steps: Based on the OpenCV algorithm, multi-modal biological feature extraction is performed on the dynamic data set, including face geometry, texture features, and iris texture features, to generate a comparison vector Bd=(Mb,Wl,Hm); At the same time, the same multi-modal biological feature extraction is performed on the security check partition face library to generate a comparison object vector Bx=(MbX,WlX,HmX); Where Mb and MbX are face geometry similarity vectors, Wl and WlX are texture feature similarity vectors, and Hm and HmX are iris feature Hamming distance vectors; Based on the comparison vector Bd=(Mb,Wl,Hm) and the comparison object vector Bx=(MbX,WlX,HmX), the difference coefficient Cy is calculated, and the specific steps are as follows: ; Among them, an amplification coefficient of a face geometry similarity vector, an amplification coefficient of a texture feature similarity vector, an amplification coefficient of an iris feature hamming distance vector; an amplification coefficient of a face geometry similarity vector an amplification coefficient of a texture feature similarity vector an amplification coefficient of an iris feature hamming distance vector , according to the discriminant ability score of each modal feature in the historical security comparison data, the amplification coefficient of each modal feature is determined according to the normalized proportion of the discriminant ability score of the modal feature and all modal features; if the collection quality of a modal feature is lower than a preset quality threshold, the amplification coefficient of the modal feature is automatically reduced, and the amplification coefficients of other modal features are proportionally increased; difference threshold values cy1 and cy2 are set, and cy1 < cy2; if Cy > cy2, the comparison result is consistent with the certificate; if cy1 < Cy ≤ cy2, the comparison result is abnormal in the comparison of the person and the certificate; if Cy ≤ cy1, the comparison result is inconsistent with the certificate; a verification unit for receiving comprehensive data and comparison results, and confirming whether the passing personnel is a missed detection user according to the comprehensive data and the comparison results, and performing secondary verification on the passing personnel when it is confirmed that there is a missed detection risk; confirming whether the passing personnel is a missed detection user according to the comprehensive data and the comparison results, and performing secondary verification on the passing personnel when it is confirmed that there is a missed detection risk, including the following steps: S601. In the isolation area, according to the missed detection coefficient in the comprehensive data, a missed detection coefficient threshold Lj is set, when the missed detection coefficient of the passing personnel ≥ Lj, the passing personnel is determined to be a high missed detection risk object, and secondary verification and correct path guidance are performed on the passing personnel; when <Lj, the passing personnel is determined to be a normal user; S602. In the security channel, when the comparison result of the passing personnel is inconsistent with the certificate, it is determined that the passing personnel has missed detection, and a re-security process is triggered; when the comparison result of the passing personnel is abnormal in the comparison of the person and the certificate, it is determined that the passing personnel has a missed detection risk, and secondary verification is performed on the passing personnel; if the secondary verification result is consistent with the certificate, the passing personnel is determined to be a normal security user; otherwise, the passing personnel is determined to have missed detection and a re-security process is triggered; when the comparison result of the passing personnel is consistent with the certificate, the passing personnel is directly determined to be a normal security user; When the comparison result of the passing person is inconsistent with the person-credential comparison result, it is determined that the passing person has walked wrong gallery bridges, and correct path guidance is performed; When the comparison result of the passing person is inconsistent with the person-credential comparison result, it is determined that the passing person has walked wrong gallery bridges, and correct path guidance is performed; When the comparison result of the passing person is inconsistent with the person-credential comparison result, it is determined that the passing person has walked wrong gallery bridges, and correct path guidance is performed;

[0012] Through intelligent behavior analysis, high-risk targets are actively identified and marked in the isolation area, and passengers entering the isolation area are accurately diverted and intensively verified to ensure that key targets are controlled. Finally, with the help of multi-modal biological feature fusion technology, continuous identity verification and dynamic tracking are performed at key nodes such as security check channels and gallery bridges. This closed-loop process not only realizes the transition from passive prevention to active intervention, but also greatly improves the real-time discovery capability and accurate interception efficiency of missed detection behavior through the layer-by-layer progressive prevention and control mechanism, while effectively avoiding flight delays caused by passengers walking to the wrong boarding gate, and comprehensively enhancing the reliability, intelligent level and comprehensive operation efficiency of the security check system.

[0013] The size of the interval and the threshold is set for easy comparison. The size of the threshold depends on the number of sample data and the base number set by the person skilled in the art for each group of sample data. As long as it does not affect the proportional relationship of the parameters and the quantized values.

[0014] The above formulas are dimensionless values calculated. The formula is obtained by software simulation of a large amount of data to obtain a formula of the nearest real situation. The preset parameters in the formula are set by the person skilled in the art according to the actual situation. The above is only the preferred specific embodiment of the present application, but the protection scope of the present application is not limited thereto. Any person skilled in the art can make equivalent substitutions or changes to the technical solutions and inventive concepts of the present application within the technical scope disclosed by the present application, which should be covered within the protection scope of the present application.

Claims

1. A security inspection and missed detection prevention system based on multimodal biometric fusion, characterized by: It includes data collection unit, risk perception unit, real-time monitoring unit, data analysis unit and verification unit; The data collection unit is used to collect the identity information and photo information of the passers-by at the security check verification counter and verify them. If the verification is passed, the passer-by's status is updated to "passed security check", and the security check status information is generated and sent to the risk perception unit. At the same time, the identity information and photo information of the passers-by are integrated to generate a security check zone face database and send it to the data analysis unit; The risk perception unit is used to receive security check status information and collect real-time video data of passers-by through high-precision wide-angle cameras in the security check channel and boarding bridge. It also completes status updates and positioning of the corresponding passers-by in the real-time video data based on the security check status information, conducts behavioral analysis on passers-by in the isolation area, calculates the missed detection coefficient of passers-by, and generates comprehensive data through integration and sends it to the verification unit. The isolation area is the restricted area from security check to the boarding gate; The real-time monitoring unit is used to collect multimodal biometric information of passers-by through smart devices deployed in security checkpoints and corridor bridge areas, integrate it to generate a dynamic data set, and send it to the data analysis unit; The data analysis unit is used to receive and process the security check zone face database and the dynamic data set, extract the multimodal biometric features of the passers-by from the dynamic data set, compare them with the corresponding features in the security check zone face database, generate the comparison results and send them to the verification unit; The verification unit is used to receive the comprehensive data and comparison results, confirm whether the passerby is a missed detection user based on the comprehensive data and comparison results, and perform a secondary verification on the passerby when it is confirmed that there is a risk of missed detection.

2. The multimodal biometric fusion security inspection and missed detection prevention system according to claim 1 is characterized in that: At the security checkpoint, the identity information and photo of the person passing through are collected and verified. The specific process is as follows: The ID card reader module built into the designated sensing area of ​​the ID card reader reads the encrypted identity information stored in the ID card of the passerby, including name, gender, ethnicity, date of birth, address, ID card number, issuing authority, validity period and standard ID photo; The built-in high-definition camera of the ID card machine detects the liveness of passers-by, including nodding, shaking, opening the mouth, and blinking. If the liveness detection is passed, a real-time frontal face photo of the passer-by is captured. The built-in face comparison algorithm of the ID card reader extracts the feature points of the real-time frontal face photo and the standard ID photo, and then compares them to calculate the similarity. Feature points include the position, outline and spacing of facial features. Set a similarity threshold Xs. If the calculated similarity score is greater than or equal to the threshold Xs, the verification is considered to be successful. Otherwise, the verification is considered to have failed and re-verification is initiated. The similarity threshold Xs is determined based on historical sample data.

3. The multimodal biometric fusion security inspection and missed detection prevention system according to claim 1 is characterized in that: Calculating the missed detection coefficient and generating annotation information includes the following steps: S301. Extracting a multimodal feature vector based on real-time video data, including the following steps: Based on the OpenPose+CNN algorithm, a multimodal feature extraction model is constructed to extract a multimodal feature vector group. The input of the multimodal feature extraction model is real-time video data, and the output is a behavior trajectory vector and an action feature vector. S302. Generate a behavior label based on the behavior trajectory vector and the action feature vector, including the dwell time, behavior switching frequency, and trailing mark; The dwell time refers to the time a user stays in a security check channel when not in a queue; The behavior switching frequency refers to the number of times a user switches behavior types within a unit of time, wherein the behavior types include standing still, walking, looking at the phone, checking signs, looking around, and looking back; The trailing marker uses a laser camera to calculate the spatial trajectory of each person in the waiting area. Through time lag synchronization analysis, if the trajectory of the later passerby is synchronized with the previous passerby’s trajectory, it is recorded as 1, otherwise it is recorded as 0; S303. Based on the behavior label, construct an abnormal behavior vector and calculate the missed detection coefficient; The abnormal behavior vector format is: U = {stop duration, behavior switching frequency, trailing mark}; The missed detection coefficient The calculation formula is: ； ； ； in, For the length of stay, is the behavior switching frequency, M is the trailing mark, is the cross-characteristic factor; is the weight coefficient, which is obtained based on historical experimental data training; Missed detection coefficient The higher the value, the greater the risk of missed detection.

4. The multimodal biometric fusion security inspection and missed detection prevention system according to claim 1, characterized in that: The smart device includes a telephoto facial camera array, an inertial measurement unit, and an edge computing gateway; the dynamic dataset includes facial geometry, texture features, and iris texture features; The edge computing gateway performs real-time preprocessing on dynamic data sets.

5. The multimodal biometric fusion security inspection and missed detection prevention system according to claim 3 is characterized in that: Generating the comparison result comprises the following steps: Based on the OpenCV algorithm, multimodal biometric features are extracted from the dynamic data set, including facial geometry, texture features, and iris texture features, to generate a comparison vector Bd=(Mb, Wl, Hm); At the same time, the same multimodal biometric feature extraction is performed on the face database of the security check area to generate the comparison object vector Bx=(MbX,WlX,HmX); Among them, Mb and MbX are facial geometric structure similarity vectors, Wl and WlX are texture feature similarity vectors, and Hm and HmX are iris feature Hamming distance vectors; Based on the comparison vector Bd=(Mb,Wl,Hm) and the comparison object vector Bx=(MbX,WlX,HmX), calculate the difference coefficient Cy. The specific steps are as follows: ; in, is the magnification factor of the facial geometric structure similarity vector, is the magnification factor of the texture feature similarity vector, is the magnification factor of the iris feature Hamming distance vector; The magnification factor of the facial geometric structure similarity vector , the magnification factor of the texture feature similarity vector and the magnification factor of the iris feature Hamming distance vector , it is set based on the discrimination ability score of each modal feature in the historical security inspection comparison data. The amplification factor of each modal feature is determined according to the normalized ratio of the discrimination ability score of this modal feature to that of all modal features. If the acquisition quality of a modal feature is lower than the preset quality threshold, the amplification factor of this modal feature is automatically reduced, and the amplification factors of other modal features are proportionally increased. Set difference thresholds cy1 and cy2, and cy1<cy2; If Cy>cy2, the comparison result is consistent with the witness; If cy1<Cy≤cy2, the comparison result is abnormal between human and evidence; If Cy≤cy1, the comparison result is inconsistent with the witness.

6. The multi-modal biometric fusion security inspection and missed detection prevention system according to claim 5, characterized in that: Based on the comprehensive data and comparison results, determine whether the passerby is a missed detection user. If it is confirmed that there is a risk of missed detection, perform a secondary verification on the passerby, including the following steps: S601. In the isolation area, based on the missed detection coefficient in the comprehensive data , set the missed detection coefficient threshold Lj, when the missed detection coefficient of the passerby When ≥Lj, the passerby is identified as a high-risk missed detection target, and the passerby is subject to secondary verification and correct route guidance; when When <Lj, the passerby is considered a normal user; S602. In the security check channel, when the person-to-person comparison result is inconsistent with the ID card, it is determined that the person has been missed and the re-inspection process is triggered; If the comparison result of the passerby's ID card is abnormal, it is determined that the passerby is at risk of being missed, and a second verification is performed on the passerby. If the second verification result is consistent with the ID card, the passerby is considered a normal security check user; otherwise, the passerby is considered to have missed the inspection and the re-inspection process is triggered. When the comparison result of the passerby is consistent with the ID card, the passerby is directly identified as a normal security check user; In the corridor passage, if the comparison result of the passerby's identity is inconsistent, it is determined that the passerby has taken the wrong corridor and will be guided to the correct path; If the comparison result of the pedestrian's ID card is abnormal, it is determined that the pedestrian is at risk of taking the wrong bridge and a second verification is performed on the pedestrian. If the second verification result is consistent with the ID card, the pedestrian is considered a normal user. Otherwise, the pedestrian is determined to have taken the wrong bridge and is guided to the correct path. When the comparison result of the passerby is consistent with the identity card, the passerby is directly identified as a normal user.

Images