A method for matching passenger information with luggage security inspection images based on artificial intelligence

Abstract

The present invention discloses a method for matching passenger information with baggage security inspection images based on artificial intelligence, which relates to the technical field of security inspection. In the present invention, a body camera and an item camera are arranged at the security inspection entrance to capture the front and side images of passengers, extract face feature vectors to generate unique identity tags, and at the same time record the external features of the baggage, and bind these features with the passenger identity tags to generate matching information; after the passengers enter the security inspection channel, the body camera uses gait recognition technology to extract the dynamic behavior trajectories of the passengers, adopts a multi-target tracking algorithm to monitor the relative positions of the passengers and the baggage in real time, and continuously updates the binding information to cope with cross scenarios; when the baggage passes through the security inspection equipment, the item camera captures X-ray images and extracts the internal features of the baggage through the YOLOv8 model, combines the internal features with the external features and the dynamic behavior trajectories, and conducts joint analysis through a multi-modal fusion algorithm. When an abnormal situation is detected, a retrospective analysis is automatically triggered.

Description

Technical Field

[0001] The present invention relates to the technical field of security inspection, and particularly to a method for matching passenger information with baggage security inspection images based on artificial intelligence. Background Art

[0002] At public transportation hubs such as airports or high-speed railway stations, the security inspection of passengers' baggage is an important link to ensure transportation safety. The method for matching passenger information with baggage mainly relies on binding the passenger identity information (such as boarding passes, identity cards) with baggage tags (such as barcodes, RFID tags). When the baggage passes through the security inspection equipment, the equipment will record its X-ray image and associate the image with the baggage tag for subsequent tracking and identification.

[0003] However, in the case of passengers' baggage being crossed or misappropriated, the recognition effect of the traditional scheme cannot meet the usage requirements. Firstly, the traditional scheme has poor adaptability to dynamic scenarios. When the baggage and the passenger are separated or cross-moved after security inspection, it is difficult to accurately track the ownership of the baggage, which easily leads to matching errors. Secondly, the traditional method highly depends on the integrity of the baggage tag. Once the tag falls off or is damaged, the matching process will be interrupted. In addition, when multiple passengers enter the security inspection area simultaneously and there is a large amount of baggage, the manual matching of baggage tags and X-ray images may be delayed or incorrect, further affecting the security inspection efficiency and accuracy.

[0004] Some solutions address the above problems by strengthening manual intervention. However, such measures not only increase the labor cost but also easily cause new problems due to communication issues. Therefore, how to improve the matching accuracy and efficiency in dynamic and complex scenarios has become an urgent problem to be solved in the current security inspection field. Summary of the Invention

[0005] In view of the above existing problems, the present invention is proposed.

[0006] The present invention provides a method for matching passenger information with baggage security inspection images based on artificial intelligence to solve the problem that the traditional scheme has poor adaptability to dynamic scenarios. When the baggage and the passenger are separated or cross-moved after security inspection, it is difficult to accurately track the ownership of the baggage, which easily leads to matching errors.

[0007] To solve the above technical problems, the present invention provides the following technical solutions:

[0008] An embodiment of the present invention provides a method for matching passenger information with baggage security inspection images based on artificial intelligence, which includes,

[0009] Step S1: Deploy a body camera and an item camera at the security checkpoint entrance. The body camera captures the frontal and side images of passengers, and combines with face recognition technology to generate a unique passenger identity label. The item camera records the external features of the luggage from a top view and binds them with the passenger identity label to generate binding information.

[0010] The external features of the luggage include geometric shape, color, and specific markings. The specific markings include iconic stickers and abnormal appearances.

[0011] Step S2: After the luggage enters the security checkpoint channel, the body camera captures the limb features and dynamic behavior trajectories of the passenger through gait recognition technology, and monitors the relative relationship between the passenger and the luggage in real time. The DeepSORT algorithm for multi-object tracking is used to track the trajectories of the passenger and the luggage, and the binding information is continuously updated.

[0012] Step S3: When the luggage passes through the security inspection equipment, the item camera captures the X-ray image, and extracts the internal features of the luggage through the YOLOv8 model. The internal features are combined with the external features in Step S1 and the behavior trajectories in Step S2 for joint analysis to determine the integrity of the matching relationship between the luggage and the passenger.

[0013] As a preferred solution of the method for matching passenger information and luggage security inspection images based on artificial intelligence according to the present invention, wherein: the step of generating a unique passenger identity label by combining face recognition technology is

[0014] Input the passenger image set I = {I f , I s}, where I f represents the frontal image of the passenger, and I s represents the side image of the passenger.

[0015] Extract the face feature vector, and the extraction formula is:

[0016] F = Softmax(W3·tanh(W2·tanh(W1·[I f , I s + b1) + b2) + b3),

[0017] where is the passenger feature vector, f j is the j-th eigenvalue, j is the feature index, tanh(x) is the hyperbolic tangent activation function, W i is the weight matrix of the i-th layer network, b i is the bias vector of the i-th layer network, d f is the dimension of the feature vector.

[0018] Generate a unique identity label ID based on the feature vector k, the generation formula is:

[0019]

[0020] Among them, ID k is the identity label of the k-th passenger, w j is the weight factor of the eigenvalue, φ(f j ) = ln(1 + |f j |) is the feature normalization function, j is the feature index, and M is the modulo operation base.

[0021] As a preferred solution of the method for matching passenger information and luggage security inspection images based on artificial intelligence according to the present invention, among them: the step of generating binding information by binding with the passenger identity label is

[0022] Input the luggage image I L , extract the external feature vector, and the extraction formula is:

[0023] L ext = [f shape (I L ), f color (I L ), f mark (I L )],

[0024] Among them,

[0025]

[0026] N c is the number of points of the luggage contour, (x i , y i ) is the coordinate of the i-th point, is the center coordinate of all feature points, α i is the weight factor of the point,

[0027]

[0028] c m is the mean value of the color components, is the overall color mean value, R, G, B are the red, green, and blue channels,

[0029]

[0030] N m is the number of pixels of the iconic sticker area, m i is the gray value of the iconic sticker,

[0031] Generate the binding matrix B k , B k = IDk L ext 。

[0032] As a preferred solution of the method for matching passenger information with luggage security inspection images based on artificial intelligence according to the present invention, wherein: the step of capturing the limb features and dynamic behavior trajectories of passengers through gait recognition technology and real-time monitoring the relative relationship between passengers and luggage is as follows:

[0033] Input the gait image sequence where G t represents the gait image at time t, and extract the gait feature vector G k , and the extraction formula is:

[0034]

[0035] wherein, is the gait feature, T f is the length of the time series, G t is the gait image, W g , b g are the network weights and biases, d g is the dimension of the gait feature vector,

[0036] Generate the dynamic behavior trajectory, and the generation formula is:

[0037]

[0038] wherein,

[0039] are the horizontal and vertical speeds of the trajectory, and t' is the time variable.

[0040] As a preferred solution of the method for matching passenger information with luggage security inspection images based on artificial intelligence according to the present invention, wherein: the step of using the multi-object tracking DeepSORT algorithm to track the trajectories of passengers and luggage and continuously update the binding information is as follows:

[0041] Define the trajectory update model, and the model formula is:

[0042]

[0043] wherein, is the updated trajectory,

[0044] Δ t is the time interval, and γ is the correction factor.

[0045] As a preferred solution of the method for matching passenger information and baggage security inspection images based on artificial intelligence according to the present invention, in which: during the joint analysis process, a global attention mechanism is introduced to optimize feature association;

[0046] During the joint analysis process, when an abnormal situation is detected, a backtracking analysis is automatically performed.

[0047] As a preferred solution of the method for matching passenger information and baggage security inspection images based on artificial intelligence according to the present invention, in which: the backtracking analysis method is as follows:

[0048] Combining historical internal features, external features, and behavior trajectories to reconfirm the attribution relationship between the baggage and the passenger;

[0049] If the abnormality cannot be resolved during the backtracking analysis, an alarm is immediately generated to prompt the security inspection personnel to intervene and perform manual confirmation.

[0050] As a preferred solution of the method for matching passenger information and baggage security inspection images based on artificial intelligence according to the present invention, in which: the behavioral actions include the actions of picking up and placing the baggage and the path trajectory;

[0051] The abnormal situations include: the separation of the trajectories of the baggage and the passenger and the passenger mistakenly taking someone else's baggage.

[0052] As a preferred solution of the method for matching passenger information and baggage security inspection images based on artificial intelligence according to the present invention, in which: the step of extracting the internal features of the baggage through the YOLOv8 model, combining the internal features with the external features in step S1 and the behavior trajectory in step S2, and performing joint analysis is as follows:

[0053] Extract the internal features of the baggage based on the X-ray image, and input the X-ray image Extract the internal features of the baggage, and the extraction formula is:

[0054]

[0055] Wherein, is the input X-ray image, representing the internal structure of the baggage, is the extracted internal feature vector of the baggage, containing d x eigenvalues, d x is the dimension of the internal feature, representing the number of features extracted from the image, (x i , y i ) is the coordinate of the i-th feature point on the image, is the central coordinate of all feature points, defined as:

[0056]

[0057] exp(x)=e x is the natural exponential function, σ is the scale parameter of the Gaussian distribution, cos(2πx i / λ) is the frequency-domain transformation of the feature points, and λ is the wavelength parameter.

[0058] As a preferred solution of the method for matching passenger information and baggage security inspection images based on artificial intelligence according to the present invention, wherein: the step of automatically performing retrospective analysis when an abnormal situation is detected is

[0059] Define the joint feature vector as U, which is expressed as:

[0060]

[0061] where U is the joint feature vector, which fuses passenger features and baggage features for matching analysis, Attention(·) is the global attention mechanism, and is defined as:

[0062]

[0063] where z represents the input feature, z i is the i-th feature component, exp(z i ) is used to calculate the weight distribution, K is the total number of passengers, and α k is the feature weight factor, representing the importance of the k-th passenger feature, and is defined as:

[0064]

[0065] F k is the feature vector of the k-th passenger, and L int , L ext represent the internal and external feature vectors of the baggage respectively;

[0066] Perform matching relationship detection, and the detection formula is:

[0067]

[0068] where S match is the matching score, used to measure the similarity between the joint feature vector U and the passenger trajectory T k , and T k =[t x,k , t y，k is the trajectory vector of the k-th passenger, where

[0069]

[0070] v x (t) is the speed of the passenger in the x direction, and v y(t) is the velocity in the y - direction,

[0071] |U| is the Euclidean norm of the joint feature vector, defined as:

[0072]

[0073] where U i is the i - th component of the joint feature vector.

[0074] The beneficial effects of the present invention are as follows: In the present invention, a body camera and an item camera are arranged at the security check entrance to capture the front and side images of passengers, extract the face feature vectors to generate unique identity tags, record the external features of the luggage at the same time, and bind these features with the passenger identity tags to generate matching information; after the passenger enters the security check channel, the body camera uses gait recognition technology to extract the dynamic behavior trajectory of the passenger, adopts a multi - target tracking algorithm to monitor the relative position of the passenger and the luggage in real - time, and continuously updates the binding information to cope with cross - scenarios; when the luggage passes through the security check equipment, the item camera captures the X - ray image and extracts the internal features of the luggage through the YOLOv8 model, combines the internal features with the external features and the dynamic behavior trajectory, and performs joint analysis through a multi - modal fusion algorithm, and at the same time introduces a global attention mechanism to optimize feature association; when an abnormal situation is detected, backtracking analysis is automatically triggered, and the attribution relationship is re - confirmed by combining historical internal features, external features and behavior trajectories.

[0075] The present invention introduces joint analysis and a global attention mechanism, which not only improves the matching efficiency, but also can accurately judge the source of abnormalities, respond quickly, and significantly enhance the accuracy and reliability of matching between passengers and luggage in dynamic scenarios. BRIEF DESCRIPTION OF THE DRAWINGS

[0076] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the following will briefly introduce the drawings required for the description of the embodiments. Obviously, the following drawings are only some embodiments of the present invention. For those of ordinary skill in the art, other drawings can be obtained based on these drawings without creative efforts.

[0077] Figure 1 is a flowchart of the method for matching passenger information and luggage security check images based on artificial intelligence of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0078] In order to make the above - mentioned objects, features and advantages of the present invention more obvious and understandable, the following will give a detailed description of the specific embodiments of the present invention with reference to the drawings in the specification.

[0079] In the following description, numerous specific details are set forth to provide a thorough understanding of the present invention. However, the present invention may be practiced in other ways than those specifically described herein. Those skilled in the art can make similar extensions without departing from the spirit of the present invention. Therefore, the present invention is not limited by the specific embodiments disclosed below.

[0080] Secondly, the so-called "one embodiment" or "embodiment" herein refers to a specific feature, structure, or characteristic that may be included in at least one implementation of the present invention. The appearances of "in one embodiment" in different places in this specification do not all refer to the same embodiment, nor are they separate or alternative embodiments that exclude each other.

[0081] Embodiment 1, referring to Figure 1 , this embodiment provides a method for matching passenger information with luggage security inspection images based on artificial intelligence, including the following steps:

[0082] Step S1, arrange a body camera and an item camera at the security inspection entrance. The body camera captures the front and side images of the passenger, and combines face recognition technology to generate a unique passenger identity label; the item camera records the external features of the luggage from a top view and binds them with the passenger identity label to generate binding information;

[0083] The external features of the luggage include geometric shape, color, and specific markings. The specific markings include iconic stickers and abnormal appearances;

[0084] The step of generating a unique passenger identity label by combining face recognition technology is

[0085] Input the passenger image set I = {I f , I s}, where I f represents the front image of the passenger, and I s represents the side image of the passenger,

[0086] Extract the face feature vector, and the extraction formula is:

[0087] F = Softmax(W3 · tanh(W2 · tanh(W1 · [I f , I s + b1) + b2) + b3),

[0088] where, is the passenger feature vector, f j is the j-th eigenvalue, j is the feature index, tanh(x) is the hyperbolic tangent activation function, W i is the weight matrix of the i-th layer network, b i is the bias vector of the i-th layer network, d f is the dimension of the feature vector,

[0089] Generate a unique identity label ID based on the feature vector k , and the generation formula is:

[0090]

[0091] where ID k is the identity label of the k-th passenger, w j is the weight factor of the eigenvalue, φ(f j ) = ln(1 + |f j |) is the feature normalization function, j is the feature index, and M is the modulus operation base;

[0092] Specifically, extract the passenger features through a three-layer neural network, and combine the normalization function and modulus operation to generate an identity label, ensuring uniqueness and distinctiveness.

[0093] The steps to generate the binding information by binding with the passenger identity label are as follows

[0094] Input the luggage image I L , extract the external feature vector, and the extraction formula is:

[0095] L ext = [f shape (I L ), f color (I L ), f mark (I L )],

[0096] where

[0097]

[0098] N c is the number of points on the luggage contour, (x i , y i ) is the coordinate of the i-th point, is the center coordinate of all feature points, α i is the weight factor of the point,

[0099]

[0100] c m is the average value of the color components, is the overall average color, R, G, B are the red, green, and blue channels,

[0101]

[0102] N m is the number of pixels in the area of the iconic sticker, m iis the grayscale value of the iconic sticker,

[0103] Generate the binding matrix B k , B k = ID k L ext ;

[0104] Specifically, through multi-scale feature extraction, obtain the external features of the luggage and bind them with the passenger identity label to generate a matrix, establishing a one-to-one correspondence between the passenger and the luggage.

[0105] Step S2: After the luggage enters the security check channel, the body camera captures the limb features and dynamic behavior trajectories of the passenger through gait recognition technology, and monitors the relative relationship between the passenger and the luggage in real time; use the multi-object tracking DeepSORT algorithm to track the trajectories of the passenger and the luggage, and continuously update the binding information;

[0106] The steps of capturing the limb features and dynamic behavior trajectories of the passenger through gait recognition technology and monitoring the relative relationship between the passenger and the luggage in real time are as follows:

[0107] Input the gait image sequence where G t represents the gait image at time t, and extract the gait feature vector G k , and the extraction formula is:

[0108]

[0109] where, is the gait feature, T f is the length of the time series, G t is the gait image, W g , b g are the network weights and biases, and d g is the dimension of the gait feature vector,

[0110] Generate the dynamic behavior trajectory, and the generation formula is:

[0111]

[0112] where,

[0113] are the horizontal and vertical speeds of the trajectory, and t' is the time variable;

[0114] Specifically, capture the dynamic behavior information of the passenger through time series feature extraction and trajectory derivation.

[0115] The steps of using the multi-object tracking DeepSORT algorithm to track the trajectories of the passenger and the luggage and continuously update the binding information are as follows:

[0116] Define the trajectory update model, and the model formula is:

[0117]

[0118] Wherein, is the updated trajectory,

[0119] Δ t is the time interval, and γ is the correction factor;

[0120] Specifically, through the trajectory correction update model, the dynamic binding of passengers and luggage is realized.

[0121] Step S3, when the luggage passes through the security inspection equipment, the item camera captures the X-ray image, and the internal features of the luggage are extracted through the YOLOv8 model. The internal features are combined with the external features in step S1 and the behavior trajectory in step S2 for joint analysis to judge the integrity of the matching relationship between the luggage and the passenger;

[0122] During the joint analysis process, the global attention mechanism is introduced to optimize the feature association;

[0123] During the joint analysis process, when an abnormal situation is detected, a backtracking analysis is automatically performed;

[0124] The backtracking analysis method is:

[0125] Combining the historical internal features, external features and behavior trajectories, reconfirm the ownership relationship between the luggage and the passenger;

[0126] If the abnormality cannot be resolved in the backtracking analysis, an alarm is immediately generated to prompt the security inspection personnel to intervene and conduct a manual confirmation;

[0127] The behavioral actions include the picking up and placing actions and path trajectories of the luggage;

[0128] The abnormal situations include: the separation of the trajectories of the luggage and the passenger and the passenger mistakenly taking someone else's luggage;

[0129] The steps of combining the internal features of the luggage extracted through the YOLOv8 model with the external features in step S1 and the behavior trajectory in step S2 for joint analysis are,

[0130] Extract the internal features of the luggage based on the X-ray image, and input the X-ray image Extract the internal features of the luggage, and the extraction formula is:

[0131]

[0132] Wherein, is the input X-ray image, representing the internal structure of the luggage, is the extracted internal feature vector of the luggage, containing dx eigenvalues, d x is the dimension of the internal feature, representing the number of features extracted from the image, (x i , y i ) are the coordinates of the i-th feature point on the image, is the central coordinate of all feature points, defined as:

[0133]

[0134] exp(x) = e x is the natural exponential function, σ is the scale parameter of the Gaussian distribution, cos(2πx i / λ) is the frequency domain transformation of the feature point, and λ is the wavelength parameter;

[0135] Specifically, by combining the Gaussian weighting in the spatial domain and the cosine transform in the frequency domain, multi-scale features of the internal structure of the luggage are extracted, enhancing the detection ability for abnormal items.

[0136] When an abnormal situation is detected, the steps for automatic backtracking analysis are as follows:

[0137] Define the joint feature vector as U, expressed as:

[0138]

[0139] where U is the joint feature vector, fusing passenger features and luggage features for matching analysis, Attention(·) is the global attention mechanism, defined as:

[0140]

[0141] where z represents the input feature, z i is the i-th feature component, exp(z i ) is used to calculate the weight distribution, K is the total number of passengers, and α k is the feature weight factor, representing the importance of the k-th passenger feature, defined as:

[0142]

[0143] F k is the feature vector of the k-th passenger, and L int , L ext represent the internal and external feature vectors of the luggage respectively;

[0144] Perform matching relationship detection, and the detection formula is:

[0145]

[0146] where S matchis the matching score, which is used to measure the similarity between the joint feature vector U and the passenger trajectory T k ; T k = [t x,k , t y，k is the trajectory vector of the k-th passenger, where

[0147]

[0148] v x (t) is the speed of the passenger in the x direction, and v y (t) is the speed in the y direction

[0149] |U| is the Euclidean norm of the joint feature vector, defined as:

[0150]

[0151] where U i is the i-th component of the joint feature vector

[0152] Specifically, the joint features are calculated through the global attention mechanism, combined with the trajectory similarity detection, to accurately match the passengers and luggage, and the abnormal situations are judged through the matching score

[0153] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and not to limit them. Although the present invention has been described in detail with reference to the preferred embodiments, those of ordinary skill in the art should understand that the technical solutions of the present invention can be modified or equivalently replaced without departing from the spirit and scope of the technical solutions of the present invention, and they should all be covered within the scope of the claims of the present invention

Claims

1. A method for matching passenger information with baggage security inspection images based on artificial intelligence, characterized in that: including, Step S1, deploy a body camera and an item camera at the security check entrance. The body camera captures the front and side images of passengers, and combines face recognition technology to generate a unique passenger identity label; The item camera records the external features of the luggage from a top view and binds them with the passenger identity label to generate binding information; The external features of the luggage include geometric shape, color, and specific markings. The specific markings include iconic stickers and abnormal appearances; Step S2, after the luggage enters the security check channel, the body camera captures the limb features and dynamic behavior trajectories of the passenger through gait recognition technology, and monitors the relative relationship between the passenger and the luggage in real time; uses the multi-object tracking DeepSORT algorithm to track the trajectories of the passenger and the luggage, and continuously updates the binding information; Step S3, when the luggage passes through the security check equipment, the item camera captures X-ray images, and extracts the internal features of the luggage through the YOLOv8 model. The internal features are combined with the external features in Step S1 and the behavior trajectories in Step S2 for joint analysis to judge the integrity of the matching relationship between the luggage and the passenger; During the process of the joint analysis, a global attention mechanism is introduced to optimize feature association; During the process of the joint analysis, when an abnormal situation is detected, automatic retrospective analysis is performed; The method of the retrospective analysis is: Combined with historical internal features, external features, and behavior trajectories, reconfirm the ownership relationship between the luggage and the passenger; If the abnormality cannot be resolved during the retrospective analysis, an alarm is immediately generated to prompt the security personnel to intervene and conduct manual confirmation; The step of automatically performing retrospective analysis when an abnormal situation is detected is, Define the joint feature vector as U, expressed as: where U is the joint feature vector, which fuses passenger features and luggage features for matching analysis, Attention(·) is the global attention mechanism, defined as: Among them, z represents the input feature, and z i is the i-th feature component, and exp(z i ) is used to calculate the weight distribution. K is the total number of passengers, and α k is the feature weight factor, representing the importance of the k-th passenger feature, and is defined as: F k is the feature vector of the k-th passenger, L int , L ext respectively represent the internal and external feature vectors of the luggage; Perform matching relationship detection, and the detection formula is: Among them, S match is the matching score, which is used to measure the similarity between the combined feature vector U and the passenger trajectory T k , and T k = [t x,k , t y,k is the trajectory vector of the k-th passenger, where v x v(t) is the velocity of the passenger in the x-direction, y v(t) is the velocity in the y-direction, |U| is the Euclidean norm of the joint feature vector, defined as: where U i is the i-th component of the combined feature vector.

2. The method for matching passenger information with baggage security inspection images based on artificial intelligence according to claim 1, characterized in that: The step of generating a unique passenger identity label by combining face recognition technology is, Input passenger image set \(I = \{I f , I s \}\), where \(I f represents the front image of the passenger, and \(I s represents the side image of the passenger. Extract the face feature vector, and the extraction formula is: F = Softmax(W3·tanh(W2·tanh(W1·[I f ,I s +b1)+b2)+b3), Among them, is the passenger feature vector, f j is the j-th eigenvalue, j is the feature index, tanh(x) is the hyperbolic tangent activation function, W i is the weight matrix of the i-th layer network, b i is the bias vector of the i-th layer network, d f is the dimension of the feature vector, Generate a unique identity tag ID based on the feature vector k , and the generation formula is: where ID k is the identity label of the k-th passenger, and w j is the weight factor of the eigenvalue, is the feature normalization function, j is the feature index, and M is the modulo operation base.

3. The method for matching passenger information with baggage security inspection images based on artificial intelligence according to claim 2, wherein: The step of binding with the passenger identity label to generate binding information is, Input luggage image I L , extract the external feature vector, and the extraction formula is: L ext = [f shape (I L ), f color (I L ), f mark (I L )], where, N c is the number of points of the luggage outline, (x i , y i ) is the coordinate of the i-th point, is the central coordinate of all feature points, α i is the weight factor of the point, c m is the mean value of color components, is the overall color mean value, and R, G, B are the red, green, and blue channels. N m is the number of pixels in the iconic sticker area, m i is the grayscale value of the iconic sticker Generate the binding matrix B k , B k = ID k L ext .

4. The method for matching passenger information with baggage security inspection images based on artificial intelligence according to claim 3, wherein: The step of capturing the limb features and dynamic behavior trajectories of the passenger through gait recognition technology and monitoring the relative relationship between the passenger and the luggage in real time is, Input gait image sequence where G t represents the gait image at time t, and extract the gait feature vector G k , and the extraction formula is: Among them, is the gait feature, T f is the time series length, G t is the gait image, W g , b g are the network weights and biases, d g is the dimension of the gait feature vector Generate the dynamic behavior trajectory, and the generation formula is: where, are the lateral and longitudinal speeds of the trajectory, and t' is the time variable.

5. The method for matching passenger information with baggage security inspection images based on artificial intelligence according to claim 4, characterized in that: The step of using the multi-object tracking DeepSORT algorithm to track the trajectories of the passenger and the luggage and continuously update the binding information is, Define the trajectory update model, and the model formula is: Among them, is the updated trajectory, Δ t is the time interval, and γ is the correction factor.

6. The method for matching passenger information with baggage security inspection images based on artificial intelligence according to claim 5, characterized in that: The behavioral actions include the picking up and placing actions and path trajectories of the luggage; The abnormal situations include: the separation of the trajectories of the luggage and the passenger and the passenger mistakenly taking someone else's luggage.

7. The method for matching passenger information with luggage security inspection images based on artificial intelligence according to claim 6, wherein: The step of extracting the internal features of the luggage through the YOLOv8 model and combining the internal features with the external features in Step S1 and the behavior trajectories in Step S2 for joint analysis is, Extract the internal features of the luggage based on the X-ray image and input the X-ray image Extract the internal features of the luggage, and the extraction formula is: Among them, is the input X-ray image, representing the internal structure of the luggage, is the extracted internal feature vector of the luggage, containing d x eigenvalues, and d x is the dimension of the internal feature, representing the number of features extracted from the image. (x i , y i ) is the coordinate of the i-th feature point on the image, is the central coordinate of all feature points, defined as: exp(x) = e x is the natural exponential function, σ is the scale parameter of the Gaussian distribution, cos(2πx i / λ) is the frequency domain transformation of the feature point, and λ is the wavelength parameter.

Images