Intelligent checking method and system for passenger certificates

Abstract

The invention discloses an intelligent checking method and system for passenger certificates, and belongs to the technical field of identity checking. According to the method, the problem that the requirement of an existing method for inspection personnel is high is solved, local feature regions in a passenger face image and a passenger certificate image are extracted through a local feature method, the local feature regions are analyzed to obtain face feature vectors in the two images and matched, the similarity between the two images is calculated through an Euclidean distance method, and the accuracy of the inspection personnel is improved. Comparing the face image with a similarity threshold value, and obtaining whether the passenger face image and the passenger certificate image are the same person according to a comparison result; compared with human eye identification, the method has the advantages that the accuracy of face image checking is improved, and identity counterfeiting is avoided; and if any one of the face feature comparison result or the text information checking result is abnormal, pausing the currently fed-back result, and preferentially feeding back the abnormal result to the checking receiving terminal, so that the timeliness of acquiring the abnormal result by the checking personnel is accelerated, and the checking efficiency of the passenger certificate is improved.

Description

Technical Field

[0001] The present invention relates to the technical field of identity verification, and in particular to an intelligent verification method and system for passenger certificates. Background Art

[0002] The necessity of checking passenger documents is to ensure the authenticity of the passenger's identity and maintain public safety and social order. It is a requirement of many laws and regulations aimed at protecting public safety and social order. By checking identity documents, public safety is guaranteed. At the same time, this also meets the needs of public security management, especially in public transportation and accommodation places, by checking identity documents, the safety of passengers is protected.

[0003] At airports and other transportation hubs, inspectors need to conduct facial comparisons and document checks to ensure that the documents held by passengers match their identities. This process requires professional training and the ability to make quick and accurate judgments from inspectors, which leads to inspectors with short years of service and insufficient experience making mistakes easily when verifying the identities of passengers, reducing the intelligence and effectiveness of passenger document verification methods.

[0004] Therefore, the existing needs are not met, and we have proposed an intelligent inspection method and system for passenger documents. Summary of the invention

[0005] The object of the present invention is to provide an intelligent inspection method and system for passenger documents, which extracts local feature areas in passenger face images and passenger document images through a local feature method, analyzes them to obtain face feature vectors in the two images and matches them, calculates the similarity between the two images through the Euclidean distance method, compares them with the similarity threshold, and determines whether the passenger face image and the passenger document image are the same person based on the comparison result; compared with human eye recognition, this method improves the accuracy of face image inspection and avoids identity fraud; if any one of the face feature comparison results or the text information inspection results is abnormal, the current result being fed back is paused, and the abnormal result is fed back to the inspection receiving terminal first, which speeds up the timeliness of the inspection personnel obtaining the abnormal result, improves the efficiency of passenger document inspection, and solves the problems raised in the above-mentioned background technology.

[0006] To achieve the above object, the present invention provides the following technical solutions:

[0007] The intelligent verification method for passenger documents comprises the following steps:

[0008] S1. Obtain text feature information and image information from the passenger's ID through a scanning device, and then obtain the passenger's facial image information through a portrait acquisition device;

[0009] S2. Extracting local feature regions of the face in the passenger face image information and the passenger ID image information respectively by using a local feature method, and analyzing the texture and structure of the local regions in the two images respectively, to obtain a passenger face feature vector and a passenger ID image face feature vector respectively;

[0010] S3, taking the passenger's facial feature vector as the vector to be compared, taking the facial feature vector in the passenger's ID image as the reference vector, matching the passenger's facial feature vector to be compared with the facial feature vector in the passenger's ID image, and calculating the similarity between the two images by using the Euclidean distance method;

[0011] S4. Preset a similarity threshold, compare the similarity between the passenger's facial feature vector and the facial feature vector in the passenger's ID image with the similarity threshold, determine whether the passenger's facial image information is the same person as the passenger's ID image information, and feed back the comparison result to the terminal held by the ID checker;

[0012] S5. Check the text feature information of the passenger's ID card online to determine whether there are any abnormalities in the cardholder information, the validity period of the card, the number of times the card is valid, the information of the accompanying person, and the verification stamp information in the passenger's ID card, and feedback the judgment result to the terminal held by the checker;

[0013] S6. The document inspector shall check the document paper information, document printing information and document binding information of the passenger's document to determine whether the production source of the passenger's document is normal, and provide the document inspector with reference according to the above inspection process.

[0014] Furthermore, after comparing the similarity between the passenger facial feature vector and the facial feature vector in the passenger ID image with the similarity threshold in S4, it is determined whether the similarity between the passenger facial feature vector of the vector to be compared and the facial feature vector in the passenger ID image of the reference vector is higher than the similarity threshold. If so, it is considered that the two faces are matched successfully and the passenger ID is the current holder; otherwise, it is considered that the match fails and the passenger ID is not the current holder.

[0015] Intelligent inspection system for passenger documents, including

[0016] Information collection unit for

[0017] The facial image of the passenger to be inspected is captured by a facial capture device, and the text information and image information in the passenger's ID are scanned by a scanning device to obtain the facial image information of the passenger to be inspected and the text feature information and image information in the passenger's ID;

[0018] Information extraction unit for

[0019] The local feature method is used to extract the local feature areas of the face in the passenger face image information and the passenger ID image information, and the local binary pattern LBP is used to analyze the texture and structure of the local area in the image to obtain the passenger face feature vector and the passenger ID image face feature vector respectively;

[0020] Similarity calculation unit, used for

[0021] Match the facial feature vector of the passenger with the facial feature vector in the passenger ID image, calculate the similarity value between the two images by using the Euclidean distance method, and compare the similarity value between the two with the preset similarity threshold to determine whether the passenger facial image information is the same person as the passenger ID image information;

[0022] Information verification unit, used for

[0023] Verify the Chinese text information on the passenger's ID card online to determine the authenticity of the passenger's identity information and the holder;

[0024] Result feedback unit for

[0025] Through wireless communication technology, the facial image comparison results and text information verification results are fed back to the verification receiving terminal in real time for reference by the document inspector;

[0026] Check receiving terminal for

[0027] After receiving the facial image comparison result and the text information verification result, the document inspector will use the two results as a basis to check the validity of the passenger's document and determine the authenticity of the current passenger's identity.

[0028] Furthermore, the information extraction unit comprises

[0029] The feature extraction module is used to extract features using the LBP algorithm and form a feature vector from the local binary pattern of each pixel;

[0030] The feature processing module is used to normalize the extracted feature vectors and eliminate the scale differences between different images;

[0031] The feature screening module is used to perform feature selection on feature vectors, select features that contribute more to the classification effect, and use a classifier to classify the feature vectors.

[0032] Furthermore, the feature extraction module includes

[0033] The sample selection module is used to select the training set and test set of previous face images, and mark the face position and face label in each sample image;

[0034] The sample processing module is used to convert the sample image into a grayscale image and perform a histogram equalization operation to improve the contrast and feature robustness of the sample image;

[0035] A face detection module is used to detect faces in images using a face detection algorithm, and to crop and normalize the face parts;

[0036] The feature dimension reduction module is used to reduce the dimension of the feature vector using the PCA principal component analysis method, thereby reducing the dimension of the feature vector and improving the separability of the feature;

[0037] The classifier training module is used to train and test the classifier using the training set and test set respectively, and evaluate the model using the accuracy, recall rate, and F1 value indicators, and to tune the model parameters.

[0038] Furthermore, the similarity calculation unit includes

[0039] A threshold setting module is used to construct a value range of the similarity parameter Sim and select a similarity threshold from the value range of the similarity parameter Sim;

[0040] The threshold comparison module is used to compare the similarity threshold with the similarity value between the passenger's facial feature vector and the facial feature vector in the passenger's ID image, and determine whether the passenger's facial image information is the same person as the passenger's ID image information based on the comparison result.

[0041] Furthermore, the information checking unit includes

[0042] The local verification module is used to first verify the identity of the cardholder and the accompanying person in the passenger's document and determine whether the current cardholder information exists within the local population information range;

[0043] The off-site verification module is used to conduct off-site online identity verification of the cardholder information and accompanying person information in the passenger's document that is not within the local population information range, and obtain the identity region of the current cardholder information.

[0044] Furthermore, the result feedback unit includes

[0045] A timing feedback module is used to feed back the facial feature comparison results or text information verification results according to the result output timing;

[0046] The priority feedback module is used to give priority feedback to the inspection receiving terminal if any abnormal state occurs in the facial feature comparison result or the text information inspection result.

[0047] Furthermore, it also includes:

[0048] Obtain the target serial number information in the passenger's ID, and obtain the static image code identification number sequence according to the target serial number information;

[0049] Determine the region to which the certificate belongs based on the sequence factor of the static image code identification number sequence, and retrieve the certificate number database of the region to which the certificate belongs;

[0050] Obtaining a stored ID code sequence of the passenger's ID name from the ID number database, and determining whether the passenger's ID is a forged ID based on the stored ID code sequence;

[0051] If not, obtain the current pixel value and grayscale value of the pixel point in the face area of ​​the passenger's ID image information, and determine the reference value weight of each pixel point according to the preset feature comparison rule;

[0052] Based on the reference value weight of each pixel point, the qualified numerical range of the pixel value and gray value of the pixel point is set;

[0053] According to the qualified value range, the current pixel value and gray value of each pixel point in the face area of ​​the passenger's ID image information are qualified;

[0054] Determine the qualification of the collected passenger document image information according to the determination result, and extract geometric feature parameters of the passenger document image information based on the qualification;

[0055] Determine defect feature distribution in geometric feature parameters of passenger document image information according to preset defect marks;

[0056] The passenger document image is reconstructed according to the defect feature distribution to obtain a reconstructed image, which is used as a reference image for facial feature extraction.

[0057] Furthermore, after collecting the facial image information, it also includes:

[0058] Obtaining face area distribution and resolution parameters based on face image information;

[0059] According to the distribution of the face area and the resolution parameters, the global face feature perception is performed through the perception loss function to obtain the perception result;

[0060] Extracting facial feature perception distribution based on perception results;

[0061] The multi-resolution spatial perceptual feature distribution of face image information is calculated based on the perceptual distribution of face features:

[0062]

[0063] Wherein, Q represents the multi-resolution spatial perception feature distribution of face image information, F represents the multi-resolution spatial perception function, S represents the facial feature perception distribution, α represents the first perception affine parameter, β represents the second perception affine parameter, μ represents the perception error factor of the multi-resolution spatial perception function, and σ represents the perception compensation factor of the multi-resolution spatial perception function;

[0064] Determine the feature extraction difficulty index corresponding to the face image information according to the multi-resolution spatial perception feature distribution;

[0065] Determine whether the facial image information collection is qualified based on the feature extraction difficulty index. If not, issue a reminder to re-collect the facial image.

[0066] Compared with the prior art, the present invention has the following beneficial effects:

[0067] 1. The present invention extracts local feature areas in the passenger face image and the passenger ID image by adopting a local feature method, and analyzes them to obtain face feature vectors in the two images; matches the face feature vectors of the two images, calculates the similarity between the two images by the Euclidean distance method, compares it with the similarity threshold, and determines whether the passenger face image and the passenger ID image are the same person based on the comparison result; thereby obtaining the face image verification result based on computer technology, and compared with human eye recognition, this method improves the accuracy of face image verification and avoids identity fraud.

[0068] 2. The present invention, through the priority feedback module, when any one of the facial feature comparison results or the text information inspection results is in an abnormal state, the result currently being fed back is suspended, and the result in the abnormal state is fed back to the inspection receiving terminal in priority; after the inspector makes a judgment on the abnormal result, he or she can choose whether to continue the result of the suspended feedback, thereby speeding up the timeliness of the inspector obtaining the abnormal result and improving the efficiency of passenger document inspection. BRIEF DESCRIPTION OF THE DRAWINGS

[0069] Figure 1 This is a flow chart of the intelligent inspection method for passenger documents of the present invention;

[0070] Figure 2 This is a diagram showing the composition of the intelligent passenger certificate inspection system of the present invention. DETAILED DESCRIPTION

[0071] 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.

[0072] In order to solve the problem that the existing inspectors need to conduct face comparison and document inspection to ensure that the documents held by passengers are consistent with their own, this process requires professional training and quick and accurate judgment of the inspectors, which leads to the fact that the inspectors with short working years and insufficient experience are prone to make mistakes when verifying the identity of passengers, which reduces the intelligence and effectiveness of the passenger document inspection method. Please refer to Figure 1-Figure 2 , this embodiment provides the following technical solutions:

[0073] In this embodiment, taking a certain station as an example, the purpose is to verify the identity of passengers entering the station to ensure the authenticity of the identity of the passengers entering the station and the safety of all passengers in the station.

[0074] The intelligent verification method for passenger documents comprises the following steps:

[0075] S1. Obtain text feature information and image information from the passenger's ID through a scanning device, and then obtain the passenger's facial image information through a portrait acquisition device; wherein the scanning device is a station inspection gate, which is used to scan the passenger's ID; the portrait acquisition device is a high-definition camera installed above the gate, which is used to obtain the real-time facial image of the current passenger, and is used to compare the image with the facial image shown in the ID to check whether the current passenger is the holder of the ID; and combined with the station's ticket purchase record, it is concluded whether the current passenger is the passenger who should enter the station to board the train.

[0076] S2. The local feature regions of the face in the passenger face image information and the passenger ID image information are extracted respectively by the local feature method, and the texture and structure of the local regions in the two images are analyzed respectively, so as to obtain the passenger face feature vector and the face feature vector in the passenger ID image respectively; specifically, the local feature method extracts the local features of the image by designing a specific feature descriptor for the face, and pre-trains and tests the classifier by collecting previous face images, and debugs and optimizes it after obtaining the test results, and after ensuring the accuracy of the classifier output results, the current passenger face image and passenger ID image are tested by the classifier respectively, and the local feature results of the two are output; secondly, because in the field of face recognition, this method pays special attention to the shape, size, position and other information of the eyes, nose, mouth and other parts, the shape, size and position information of the eyes, nose and mouth in the face image can be used as the face feature vector in the two images.

[0077] S3. Use the passenger's facial feature vector as the vector to be compared, use the facial feature vector in the passenger ID image as the reference vector, match the passenger's facial feature vector to be compared with the facial feature vector in the passenger ID image, and calculate the similarity between the two images using the Euclidean distance method. Specifically, by matching the passenger's facial image information with the shape, size, and position information of the eyes, nose, and mouth in the passenger ID image information, the Euclidean distance method calculates the Euclidean distance based on the direct distance between the feature vectors in the two images. The size of the Euclidean distance is inversely proportional to the similarity between the images, that is, the greater the distance, the smaller the similarity; the smaller the distance, the greater the similarity; thus, the similarity between the two images is evaluated through the calculation result.

[0078] S4. Preset a similarity threshold, compare the similarity between the passenger's facial feature vector and the facial feature vector in the passenger's ID image with the similarity threshold, and determine whether the passenger's facial image information is the same person as the passenger's ID image information, and feed back the comparison result to the terminal held by the ID inspection personnel; after comparing the similarity between the passenger's facial feature vector and the facial feature vector in the passenger's ID image with the similarity threshold, determine whether the similarity between the passenger's facial feature vector of the to-be-compared vector and the facial feature vector in the passenger's ID image of the reference vector is higher than the similarity threshold; if higher than the similarity threshold, it is considered that the two faces match successfully, and the passenger ID is the current ID holder; otherwise, it is considered that the match fails, and the passenger ID is not the current ID holder;

[0079] S5. Conduct online verification of the text feature information in the passenger's ID, determine whether there are any abnormalities in the ID holder information, ID validity period, ID validity times, accompanying person information, and verification stamp information in the passenger's ID, and feedback the judgment result to the terminal held by the verification personnel; specifically, when verifying the passenger's ID, the ID holder information is preferably checked within the scope of local population information, thereby narrowing the verification scope and simplifying the identity verification process; if the current ID holder information is not within the scope of local population information, conduct an off-site online verification, thereby reducing the pressure on the off-site online verification channel and optimizing the diversity of passenger document verification.

[0080] S6. The document inspector shall check the document paper information, document printing information and document binding information of the passenger's document to determine whether the production source of the passenger's document is normal, and provide the document inspector with reference according to the above inspection process.

[0081] Intelligent inspection system for passenger documents, including:

[0082] Information collection unit for

[0083] The facial image of the passenger to be inspected is captured by a portrait acquisition device, and the text information and image information in the passenger's ID are scanned by a scanning device to obtain the facial image information of the passenger to be inspected and the text feature information and image information in the passenger's ID. Specifically, the scanning device is a station inspection gate, which is designed to scan the text information and image information in the passenger's ID card; the portrait acquisition device is a high-definition camera installed above the gate, which is designed to obtain the real-time facial image of the current passenger.

[0084] Information extraction unit for

[0085] The local feature region of the face in the passenger face image information and the passenger ID image information is extracted by the local feature method, and the texture and structure of the local region in the image are analyzed by the local binary pattern LBP to obtain the passenger face feature vector and the face feature vector in the passenger ID image respectively; the information extraction unit includes:

[0086] The feature extraction module is used to extract features using the LBP algorithm and form a feature vector from the local binary pattern of each pixel; the feature extraction module includes:

[0087] The sample selection module is used to select training sets and test sets of previous face images, and mark the face position and face label in each sample image; specifically, by collecting multiple groups of face images in the past as sample sets, and selecting clearer face images among them for classification, they are divided into training sets and test sets. The training set aims to determine the feature points of the classifier, and the test set aims to determine whether the training results of the classifier are accurate, thereby optimizing the training effect of the classifier.

[0088] The sample processing module is used to convert the sample image into a grayscale image and perform a histogram equalization operation to improve the contrast of the sample image and the robustness of the features. Specifically, histogram equalization is an image processing technology that expands the brightness range by redistributing the pixel values ​​of the face image, thereby making the brightness distribution of the face image more uniform. Specifically, the grayscale histogram of the face image is determined, that is, the face image is converted into a grayscale face image, and the number of pixels at each grayscale level is calculated; the cumulative distribution function is calculated, the grayscale histogram is normalized, and the cumulative probability distribution of the pixel values ​​is obtained; the mapping function is calculated, and the new pixel value corresponding to each grayscale level is calculated according to the cumulative distribution function value; the pixel value is remapped, and each pixel value of the original face image is replaced by the corresponding new pixel value according to the mapping function, thereby improving the contrast of the sample image and the robustness of the features.

[0089] The face detection module is used to detect faces in images using a face detection algorithm, and to crop and standardize the face part. Specifically, in the face images collected in the past, due to shooting techniques or angle factors, the entire image often contains redundant background images in addition to face information. Therefore, in order to ensure the accuracy of classifier training, the sample images need to be cropped to reduce the proportion of background images in the sample images and ensure that the face information is in the main position of the overall image.

[0090] The feature dimension reduction module is used to perform dimension reduction operations on feature vectors using the PCA principal component analysis method, thereby reducing the dimension of the feature vector and improving the separability of the features. Specifically, the PCA principal component analysis method is a supervised dimension reduction method that aims to maximize the distance between different categories and minimize the distance between the same categories to achieve better classification results.

[0091] The classifier training module is used to train and test the classifier using the training set and the test set respectively, and use the accuracy, recall rate, and F1 value indicators to evaluate the model and tune the model parameters; specifically, the above-processed training set and test set can be imported into the classifier in turn for training and testing to improve the classification effect of the classifier, so as to facilitate the classification and extraction of feature information of the current passenger's facial image and the image information in the passenger's ID.

[0092] The feature processing module is used to normalize the extracted feature vectors to eliminate the scale differences between different images. Specifically, the purpose of normalization is to improve the comparability of sample sets and the performance of classifiers. By adjusting the values ​​of different features to the same scale, it is possible to avoid features with a larger numerical range from having too much influence in classifier training, thereby making the classifier more robust and reliable. In addition, normalization also helps to accelerate the convergence speed of the classifier training process.

[0093] The feature screening module is used to perform feature selection on the feature vector, select the features that contribute more to the classification effect, and use the classifier to classify the feature vector; specifically, as mentioned above, in the field of face recognition, special attention is paid to the shape, size, position and other information of the eyes, nose, mouth and other parts; therefore, the shape, size and position information of the eyes, nose and mouth in the face image can be used as the face feature vector in both images; after the face feature vector is selected, the feature vector is classified by the classifier to improve the classification precision of the classifier, thereby further facilitating the inspection of the face features in the two images.

[0094] Similarity calculation unit, used for

[0095] Match the facial feature vector of the passenger with the facial feature vector in the passenger ID image, calculate the similarity value between the two images by the Euclidean distance method, and compare the similarity value between the two with a preset similarity threshold to determine whether the passenger facial image information and the passenger ID image information are the same person; the similarity calculation unit includes:

[0096] The threshold setting module is used to construct a value range of the similarity parameter Sim and select a similarity threshold from the value range of the similarity parameter Sim; specifically, it is usually represented by the parameter Sim, whose value range is 0.00 to 100.00, and the recommended threshold is 80. When the recommended similarity is ≥70, it can be judged as the same person.

[0097] The threshold comparison module is used to compare the similarity threshold with the similarity value between the passenger's facial feature vector and the facial feature vector in the passenger's ID image, and determine whether the passenger's facial image information is the same person as the passenger's ID image information based on the comparison result; specifically, it can be seen from the above that: after the similarity value between the two images is calculated by the Euclidean distance method, when the similarity value between the two images is ≥70, it means that the current passenger identity and the identity of the document holder are the same person; otherwise, the opposite is true; based on this, the passenger's ID and facial image inspection results are obtained according to computer technology, avoiding the disadvantages of manual naked eye recognition.

[0098] Information verification unit, used for

[0099] The Chinese text information of the passenger's certificate is checked online to determine the authenticity of the passenger's identity information and the holder; the information checking unit includes:

[0100] The local verification module is used to prioritize local identity verification of the cardholder information and accompanying person information in the passenger's certificate to determine whether the current cardholder information exists within the local population information range. Specifically, when verifying the text information in the passenger's certificate, priority can be given to checking the local population information range. Since the local population information query is not affected by various cross-regional permissions, it is prioritized for local query, thereby improving the timeliness of identity verification for some local passengers.

[0101] The off-site verification module is used to conduct off-site online identity verification of the cardholder information and accompanying person information in the passenger documents that are not within the scope of the local population information, and obtain the identity region of the current cardholder information; specifically, off-site online query is a necessary and unobtainable step; when checking the passenger documents, if the current passenger does not exist in the local population information range, it is necessary to conduct an off-site verification through the Internet in order to know the current passenger's identity region, so as to judge the security of the current passenger's identity; at the same time, combined with the station's internal ticket purchasing system, it is known whether the current passenger is a passenger who should enter the station, thereby improving the practicality of passenger document verification and the safety of passengers in the station.

[0102] Result feedback unit for

[0103] The face image comparison result and the text information verification result are respectively fed back to the verification receiving terminal in real time through wireless communication technology for reference by the document inspector; the result feedback unit includes:

[0104] The timing feedback module is used to feed back the facial feature comparison results or text information verification results according to the result output timing; specifically, during the operation of the passenger document inspection system, it is limited by the actual network speed; therefore, the system has different efficiencies in facial feature inspection and document text inspection; in order to avoid congestion in the network channel, when the data results are fed back, they can be transmitted and fed back in the order in which the data results are obtained, thereby optimizing the performance of the transmission channel and avoiding the phenomenon of poor network or data transmission failure when two results are transmitted at the same time.

[0105] The priority feedback module is used to give priority to feeding back to the inspection receiving terminal the abnormal status of either the facial feature comparison result or the text information inspection result; specifically, during the transmission of a certain result, such as: when the facial feature comparison result is transmitted, the passenger's ID is found to be expired during the text information inspection of the passenger's ID. At this time, the transmission path of the facial feature comparison result is automatically suspended, and the abnormal information that the current passenger's ID has expired is reported first, thereby speeding up the efficiency of the inspection personnel in obtaining the abnormal status of the passenger's ID, so that the inspection personnel can quickly solve the corresponding problems for the current passenger according to the factors of the abnormal status, thereby optimizing the practicality of the passenger document inspection system.

[0106] Check receiving terminal for

[0107] After receiving the facial image comparison result and the text information verification result, the document inspector will use the two results as a basis to check the validity of the passenger's document and determine the authenticity of the current passenger's identity.

[0108] The beneficial effects achieved by the above content are: through the above method, the face image verification results are obtained based on computer technology. Compared with human eye recognition, this method improves the accuracy of face image verification and avoids identity fraud; at the same time, it speeds up the timeliness of the inspection personnel obtaining abnormal results and improves the efficiency of passenger document inspection.

[0109] Working principle: The local feature method is used to extract the local feature areas in the passenger's facial image and the passenger's ID image, and the facial feature vectors in the two images are obtained by analyzing them; the facial feature vectors of the two images are matched, and the similarity between the two images is calculated by the Euclidean distance method, and it is compared with the similarity threshold. According to the comparison result, it is concluded whether the passenger's facial image and the passenger's ID image are the same person; through the priority feedback module, when either the facial feature comparison result or the text information inspection result is abnormal, the current feedback result will be paused, and the result of the abnormal state will be fed back to the inspection receiving terminal first.

[0110] In one embodiment, it further includes:

[0111] Obtain the target serial number information in the passenger's ID, and obtain the static image code identification number sequence according to the target serial number information;

[0112] Determine the region to which the certificate belongs based on the sequence factor of the static image code identification number sequence, and retrieve the certificate number database of the region to which the certificate belongs;

[0113] Obtaining a stored ID code sequence of the passenger's ID name from the ID number database, and determining whether the passenger's ID is a forged ID based on the stored ID code sequence;

[0114] If not, obtain the current pixel value and grayscale value of the pixel point in the face area of ​​the passenger's ID image information, and determine the reference value weight of each pixel point according to the preset feature comparison rule;

[0115] Based on the reference value weight of each pixel point, the qualified numerical range of the pixel value and gray value of the pixel point is set;

[0116] According to the qualified value range, the current pixel value and gray value of each pixel point in the face area of ​​the passenger's ID image information are qualified;

[0117] Determine the qualification of the collected passenger document image information according to the determination result, and extract geometric feature parameters of the passenger document image information based on the qualification;

[0118] Determine defect feature distribution in geometric feature parameters of passenger document image information according to preset defect marks;

[0119] The passenger document image is reconstructed according to the defect feature distribution to obtain a reconstructed image, which is used as a reference image for facial feature extraction.

[0120] The beneficial effects of the above technical solution are: it can accurately determine whether the passenger's document is a forged document and thus judge the passenger's identity security, thereby improving security. Furthermore, by evaluating the defect features of the passenger's document image information and then reconstructing the image, the high quality of the image and the data reference significance can be guaranteed, laying the foundation for the subsequent global extraction of facial features, further improving practicality.

[0121] In one embodiment, after collecting the facial image information, the method further includes:

[0122] Obtaining face area distribution and resolution parameters based on face image information;

[0123] According to the distribution of the face area and the resolution parameters, the global face feature perception is performed through the perception loss function to obtain the perception result;

[0124] Extracting facial feature perception distribution based on perception results;

[0125] The multi-resolution spatial perceptual feature distribution of face image information is calculated based on the perceptual distribution of face features:

[0126]

[0127] Wherein, Q represents the multi-resolution spatial perception feature distribution of face image information, F represents the multi-resolution spatial perception function, S represents the facial feature perception distribution, α represents the first perception affine parameter, β represents the second perception affine parameter, μ represents the perception error factor of the multi-resolution spatial perception function, and σ represents the perception compensation factor of the multi-resolution spatial perception function;

[0128] Determine the feature extraction difficulty index corresponding to the face image information according to the multi-resolution spatial perception feature distribution;

[0129] Determine whether the facial image information collection is qualified based on the feature extraction difficulty index. If not, issue a reminder to re-collect the facial image.

[0130] The beneficial effect of the above technical solution is: by calculating the multi-resolution spatial perception feature distribution of facial image information, the qualifiedness of the collected facial images can be judged based on the dimensions of the multi-resolution mapping space to determine whether they are qualified, thereby ensuring the qualification of the collected facial images and their compliance with requirements, and improving practicality and data quality.

[0131] It should be noted that, in this article, relational terms such as first and second, etc. are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Moreover, the terms "include", "comprise" or any other variants thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or device.

[0132] Although embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions and variations may be made to the embodiments without departing from the principles and spirit of the present invention, and that the scope of the present invention is defined by the appended claims and their equivalents.

Claims

1. An intelligent inspection method for passenger documents, characterized in that: The following steps are involved: S1. Obtain text feature information and image information from the passenger's ID through a scanning device, and then obtain the passenger's facial image information through a portrait acquisition device; S2. Extracting local feature regions of the face in the passenger face image information and the passenger ID image information respectively by using a local feature method, and analyzing the texture and structure of the local regions in the two images respectively, to obtain a passenger face feature vector and a passenger ID image face feature vector respectively; S3, taking the passenger's facial feature vector as the vector to be compared, taking the facial feature vector in the passenger's ID image as the reference vector, matching the passenger's facial feature vector to be compared with the facial feature vector in the passenger's ID image, and calculating the similarity between the two images by using the Euclidean distance method; S4. Preset a similarity threshold, compare the similarity between the passenger's facial feature vector and the facial feature vector in the passenger's ID image with the similarity threshold, determine whether the passenger's facial image information is the same person as the passenger's ID image information, and feed back the comparison result to the terminal held by the ID checker; S5. Check the text feature information of the passenger's ID card online to determine whether there are any abnormalities in the cardholder information, the validity period of the card, the number of times the card is valid, the information of the accompanying person, and the verification stamp information in the passenger's ID card, and feedback the judgment result to the terminal held by the checker; S6. The document inspector shall check the document paper information, document printing information and document binding information of the passenger's document to determine whether the production source of the passenger's document is normal and provide it for reference by the document inspector according to the inspection process.

2. The intelligent inspection method for passenger documents according to claim 1, characterized in that: After comparing the similarity between the passenger's facial feature vector and the facial feature vector in the passenger's ID image with the similarity threshold in S4, it is determined whether the similarity between the passenger's facial feature vector of the to-be-compared vector and the facial feature vector in the passenger's ID image of the reference vector is higher than the similarity threshold. If so, it is considered that the two faces are matched successfully and the passenger's ID is the current holder; if so, it is considered that the match fails and the passenger's ID is not the current holder.

3. An intelligent inspection system for passenger documents, applied to the intelligent inspection method for passenger documents according to any one of claims 1 to 2, characterized in that: include Information collection unit for The facial image of the passenger to be inspected is captured by a facial capture device, and the text information and image information in the passenger's ID are scanned by a scanning device to obtain the facial image information of the passenger to be inspected and the text feature information and image information in the passenger's ID; Information extraction unit for The local feature method is used to extract the local feature areas of the face in the passenger face image information and the passenger ID image information, and the local binary pattern LBP is used to analyze the texture and structure of the local area in the image to obtain the passenger face feature vector and the passenger ID image face feature vector respectively; Similarity calculation unit, used for Match the facial feature vector of the passenger with the facial feature vector in the passenger ID image, calculate the similarity value between the two images by using the Euclidean distance method, and compare the similarity value between the two with the preset similarity threshold to determine whether the passenger facial image information is the same person as the passenger ID image information; Information verification unit, used for Verify the Chinese text information on the passenger's ID card online to determine the authenticity of the passenger's identity information and the holder; Result feedback unit for Through wireless communication technology, the facial image comparison results and text information verification results are fed back to the verification receiving terminal in real time for reference by the document inspector; Check receiving terminal for After receiving the facial image comparison result and the text information verification result, the document inspector will use the two results as a basis to check the validity of the passenger's document and determine the authenticity of the current passenger's identity.

4. The intelligent passenger document checking system according to claim 3, characterized in that: The information extraction unit comprises The feature extraction module is used to extract features using the LBP algorithm and form a feature vector from the local binary pattern of each pixel; The feature processing module is used to normalize the extracted feature vectors and eliminate the scale differences between different images; The feature screening module is used to perform feature selection on feature vectors, select features that contribute more to the classification effect, and use a classifier to classify the feature vectors.

5. The intelligent passenger document checking system according to claim 4, characterized in that: The feature extraction module includes The sample selection module is used to select the training set and test set of previous face images, and mark the face position and face label in each sample image; The sample processing module is used to convert the sample image into a grayscale image and perform a histogram equalization operation to improve the contrast and feature robustness of the sample image; A face detection module is used to detect faces in images using a face detection algorithm, and to crop and normalize the face parts; The feature dimension reduction module is used to reduce the dimension of the feature vector using the PCA principal component analysis method, thereby reducing the dimension of the feature vector and improving the separability of the feature; The classifier training module is used to train and test the classifier using the training set and test set respectively, and evaluate the model using the accuracy, recall rate, and F1 value indicators, and to tune the model parameters.

6. The intelligent passenger document checking system according to claim 3, characterized in that: The similarity calculation unit comprises A threshold setting module is used to construct a value range of the similarity parameter Sim and select a similarity threshold from the value range of the similarity parameter Sim; The threshold comparison module is used to compare the similarity threshold with the similarity value between the passenger's facial feature vector and the facial feature vector in the passenger's ID image, and determine whether the passenger's facial image information is the same person as the passenger's ID image information based on the comparison result.

7. The intelligent passenger document checking system according to claim 3, characterized in that: The information checking unit comprises The local verification module is used to first verify the identity of the cardholder and the accompanying person in the passenger's document and determine whether the current cardholder information exists within the local population information range; The off-site verification module is used to conduct off-site online identity verification of the cardholder information and accompanying person information in the passenger's document that is not within the local population information range, and obtain the identity region of the current cardholder information.

8. The intelligent passenger document checking system according to claim 3, characterized in that: The result feedback unit comprises A timing feedback module is used to feed back the facial feature comparison results or text information verification results according to the result output timing; The priority feedback module is used to give priority feedback to the inspection receiving terminal if any abnormal state occurs in the facial feature comparison result or the text information inspection result.

9. The intelligent passenger document checking system according to claim 1, characterized in that: Also includes: Obtain the target serial number information in the passenger's ID, and obtain the static image code identification number sequence according to the target serial number information; Determine the region to which the certificate belongs based on the sequence factor of the static image code identification number sequence, and retrieve the certificate number database of the region to which the certificate belongs; Obtaining a stored ID code sequence of the passenger's ID name from the ID number database, and determining whether the passenger's ID is a forged ID based on the stored ID code sequence; If not, obtain the current pixel value and grayscale value of the pixel point in the face area of ​​the passenger's ID image information, and determine the reference value weight of each pixel point according to the preset feature comparison rule; Based on the reference value weight of each pixel point, the qualified numerical range of the pixel value and gray value of the pixel point is set; According to the qualified value range, the current pixel value and gray value of each pixel point in the face area of ​​the passenger's ID image information are qualified; Determine the qualification of the collected passenger document image information according to the determination result, and extract geometric feature parameters of the passenger document image information based on the qualification; Determine defect feature distribution in geometric feature parameters of passenger document image information according to preset defect marks; The passenger document image is reconstructed according to the defect feature distribution to obtain a reconstructed image, which is used as a reference image for facial feature extraction.

10. The intelligent passenger document checking system according to claim 1, characterized in that: After collecting the facial image information, it also includes: Obtaining face area distribution and resolution parameters based on face image information; According to the distribution of the face area and the resolution parameters, the global face feature perception is performed through the perception loss function to obtain the perception result; Extracting facial feature perception distribution based on perception results; The multi-resolution spatial perceptual feature distribution of face image information is calculated based on the perceptual distribution of face features: Wherein, Q represents the multi-resolution spatial perception feature distribution of face image information, F represents the multi-resolution spatial perception function, S represents the facial feature perception distribution, α represents the first perception affine parameter, β represents the second perception affine parameter, μ represents the perception error factor of the multi-resolution spatial perception function, and σ represents the perception compensation factor of the multi-resolution spatial perception function; Determine the feature extraction difficulty index corresponding to the face image information according to the multi-resolution spatial perception feature distribution; Determine whether the facial image information collection is qualified based on the feature extraction difficulty index. If not, issue a reminder to re-collect the facial image.

Images